Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
APPLICATION OF REMOTE SENSING TO HYDROLOGY
FINAL TECHNICAL REPORT
AQS1A-CR-120278) APPLICATION OF RENOESE1SInG TO HYDROLOGY Final Report 7-81International Business nachines Corp.)
CSCL 08 Unclas
G3/13 42714
PRICES SUBJECT TO CHANGE
Reproduced by
NATIONAL TECHNICALINFORMATION SERVICE
US Department of CommerceSpringfield, VA. 22151-
https://ntrs.nasa.gov/search.jsp?R=19740019698 2020-06-26T19:16:54+00:00Z
APPLICATION OF REMOTE SENSING TO HYDROLOGY
FINAL TECHNICAL REPORT
IBM No. 73W-00387
September 1973
Prepared for:National Aeronautics and Space AdministrationGeorge C. Marshall Space Flight Center
Under Contract NAS8-14000/SA 2171Technical Directives IU-25-72 and IU-321-73MSFC DR L 008-A, Line Item 161
- Federal Systems Division, Electronics Systems Center/Huntsville, Alabama
PREFACE
This report is expected to receive wider distribution than anypublished previously on the same study. It therefore incorporates muchof the material contained in those prior reports, for the convenience ofreaders who have not had access to them. Those already familiar withthe study concept and prior progress are entreated to understand andforgive the duplications.
The study is the first part (a feasibility investigation) of a longer-range application development project whose concept has evolved from anidea puggested at the International Astronautical Congress in Brussels in1971 . Since it was initiated under NASA contract in July 1971, it hasfocused on the same objective even though (as in all such research) thetechnical scope, methodology and manpower levels have from time to timebeen revised. A brief history of the study appears in Appendix A.
The study objective and concept are described in Section 1, with asummary of results. Section 2 supports the choice of simulation modeland describes its use in the study. Section 3 presents the study methodologyand describes each task and the results of each task. Recommendations foradditional research and validation efforts appear in Section 4. Appendicescontain a summary project history, a list of references and the basicdistribution list for the report.
The authors are grateful to many individuals for initiation and supportof and contributions to the study. Dr. Peter A. Castruccio (presentlypresident of Ecosystems International, Inc.) and Mr. Andrew Adelman of IBMwere energetic in promoting the basic idea. Dr. George McDonough ofNASA's George C. Marshall Space Flight Center sponsored the projectand was instrumental in securing funds to support it. Dr. William D. Clarke,as NASA-MSFC technical manager, made several helpful suggestions andfirst recognized the potential value to others of the processed and integrateddata base which is a by-product of the study.
The advice of authorities in many areas related to streamflow fore-casting and hydrologic modeling was available to the study team.Dr. L. Douglas James of the Environmental Research Center, GeorgiaInstitute of Technology, provided the basic simulation and calibrationprograms, advised us in their implementation and reviewed interim resultsof the study. In obtaining historical and physiographic data, selecting
watersheds for study, and understanding the objectives and methods ofoperational river forecasting, the authors have received invaluableassistance from personnel of the Tennessee Valley Authority; NationalWeather Service Hydrology Office and Lower Mississippi River ForecastCenter; U. S. Geological Survey, in several locations; U. S. Soil Conser-vation Service; and the U. S. Army Corps of Engineers.
Mrs. Becky Robinson of IBM deserves a special measure of thanks.In a remarkable exhibition of versatility, in addition to performing asdepartmental secretary, she:
* secured storage facilities for the tons of data acquiredand generated in the study;
* constructed photomosaics and overlay graphics;
* identified and maintained hundreds of maps and aerialphotographs;
* made special trips to TVA in Knoxville to obtainhistorical data;
* prepared illustrations for presentations and reports;
* typed, edited and assembled technical reports, includingthis one.
These acknowledgements of assistance do not imply any endorsementor approval by those whose help we appreciate so much. The conduct ofthe study, the results achieved and the content of this report are theresponsibility of the authors. Any errors of fact, methodology or logicshould be attributed solely to us.
Reuben Ambaruch JohnW. SimmonsPrincipal Investigator Study Manager
ii
CONTENTS
Section Page
Preface i
List of Figures vi
List of Tables viii
1 INTRODUCTION 1-1
1. 1 Objective 1-11. 2 Basin Modeling and Streamflow Forecasting 1-11. 3 Potential Application of Remote Sensing 1-31. 4 Feasibility Study Methodology 1-41.4.1 Principal Study Elements 1-41.4.2 Calibration 1-91. 4. 3 Correlation 1-91.4.4 Validation 1-91.5 Summary of Results 1-91. 5. 1 Feasibility Clearly Indicated 1-91. 5.2 Significant By-Products 1-101.5.2.1 Integrated Data Base 1-101.5.2.2 The IBM System for Simulation and
Analysis of Watersheds 1-111.6 A Glimpse of the Future 1-11
2 SELECTION OF SIMULATION MODEL 2-1
2. 1 Objective 2-12.2 Stochastic Models 2-12.3 Parametric Models 2-12. 4 Stanford Watershed Model (SWM) 2-22. 5 Kentucky Watershed Model (KWM) and
OPSET Program 2-32. 5. 1 Kentucky Watershed Model 2-52. 5.2 OPSET 2-52. 6 Modifications and Additions to KWM and
OPSET 2-9
3 TECHNICAL PROGRESS 3-1
3. 1 Physiographic Data Base 3-43.1.1 Area 3-43. 1.2 Overland Flow Surface Length (OFSL) 3-4
iii
CONTENTS (Continued)
Section Page
3.1.3 Overland Flow Surface Slope (OFSS) 3-93.1.4 Time Area Histogram 3-93.1.5 Impervious Surface Fraction (FIMP) 3-123.1.6 Water Surface Fraction (FWTR) 3-163.1.7 Vegetative Interception Maximum Rate
(VINTMR) 3-163.1.8 Overland Flow Manning's "N" for
Previous Surfaces (OFMN) 3-163.1.9 Overland Flow Manning's "N" for
Impervious Surfaces (OFMNIS) 3-163.1.10 Ground Water Evapotranspiration
Factor (GWETF) 3-163.1.11 Subsurface Water Factor (SUBWF) 3-183.2 Historical Data Base 3-183.2.1 Data Availability Summary 3-223.2.2 Hourly Precipitation Data 3-243.2.3 Selected Storm Data 3-253.2.4 Daily Discharge Data 3-263.2.5 Evaporation Tables 3-263. 3 Master Watershed Data Bank 3-263.4 Calibration (Task 4) 3-283.4.1 Use of OPSET 3-283.4.2 Parameters Estimated by Calibration 3-303.4.3 Calibration Results 3-323. 5 Sensitivity Analysis (Task 5) 3-433.6 Correlation (Task 6) 3-533.7 Validation (Task 7) 3-573.8 Resolution/Accuracy Analysis
(Task 8) 3-57
4 RECOMMENDED ADDITIONAL RESEARCH 4-1
4.1 Expanded Correlation Data Base 4-14.2 Refined Correlation Study 4-14. 3 Validate Technique in Other Regions 4-24.4 Document the System for Simulation
and Analysis of Watersheds 4-24.5 Pilot Demonstration, Operational
River Forecast Application 4-2
iv
CONTENTS (Continued)
Section Page
APPENDIX A SUMMARY OF STUDY HISTORY A-i
APPENDIX B REFERENCES B-i
APPENDIX C PROGRESS REPORT BASIC C-iDISTRIBUTION LIST
V
LIST OF FIGURES
Figure Page
1. Cross Section of Idealized Rural Catchment 1-2
2. Study Methodology 1-5
3. Location of Initially Chosen Watersheds 1-8
4. Moisture Accounting in the Stanford Watershed Model 2-4
5. Simulation Model (KWM) Inputs and Output 2-6
6. Parameter Optimization Concept 2-7
7. Calibration Program (OPSET) Inputs and Outputs 2-8
8. Example of Yearly Statistical Summary 2-11
9. Example of Multi-Year Statistical Summary 2-12
10. Example of Output Summary 2-13
11. Example of Observed Hydrograph Data 2-15
12. Typical Stream Gage Rating Table 2-16
13. Feasibility Study Flow 3-2
14. Physiographic Data Base 3-5
15. Time Area Histogram Calculation (Typical) 3-13
16. Typical Watershed (Noland Creek, N. C.) Dividedby Isochrones 3-14
17. Task 2: Historical Data Base 3-19
18. Historical Data Availability 3-23
19. Task 3: Master Watershed Data Bank 3-27
20. Task 4: Calibration 3-29
21. One Year Simulated and Observed Flow, WhiteHollow, 1961 3-37
22. One Month Simulated and Observed Flow, WhiteHollow, March 1961 3-38
23. Selected Storm, November 16-23, 1957, WhiteHollow 3-39
24. One Month Simulated and Observed Flows, LittleChestuee, March 1951 3-40
25. Simulated and Observed Runoff, Major Winter StormLittle Chestuee, Mar. 28-29, 1951 3-41
26. Simulated and Observed Flows, Emory River,February 1956 3-42
vi
LIST OF FIGURES (Continued)
Figure Page
27. Task 5: Sensitivity Analysis 3-45
28. Task 6: Correlation 3-54
29. Task 7: Validation 3-58
30. Example of Mean Daily Streamflow Plots 3-60
31. Example of Hourly Streamflow Plots 3-61
32. Ultimate Method 3-62
vii
LIST OF TABLES
Table Page
1. Selected Watersheds 1-7
2. "Raw" Physiographic Data Acquired(Tennessee Valley Region) 3-6
3. Physiographic Data Base Status 3-7
4. Physiographic Data Parameters 3-8
5. Worksheet for OFSS/OFSL (Typical) 3-10
6. Time Area Histogram Worksheet (Typical) 3-15
7. Manning's Roughness Coefficient for OverlandFlow for Various Surface Types 3-17
8. "Raw" Historical Data Acquired(Tennessee Valley Region) 3-20
9. Processed Historical Data 3-21
10. Simulation Model Parameters After Calibration 3-33
11. Model Parameter Definitions 3-34
12. Simulation Accuracy After Calibration 3-35
13. Sensitivity Analysis of Little Chestuee 3-46
14. Summary of Sensitivity Analysis for Little Chestuee 3-50
15. Correlation Results Multiple Regression Analysis 3-55
16. Statistical Characteristics, Regression Equations 3-56
17. Validation Results 3-59
viii
SECTION 1
INTRODUCTION
1.1 OBJECTIVE
The objective of the study whose results are reported herein is to
assess the feasibility of using the data produced by remote observation from
space and/or aircraft to reduce the time and expense normally involved in
achieving the ability to predict the hydrological behavior of an ungagedwatershed. Such a capability will enhance effective planning for urban and
industrial development, flood control, hydroelectric power, navigation,and water resources management. Traditionally, the ability to predictstreamflow rates, in response to given precipitation, is attainable only after
installation of extensive instrumentation and collection of data for several
years to estimate the operating parameters of a model of the basin or water-shed system. This project was aimed at developing techniques for using the
large amounts of data acquired by earth observation missions to determinethose parameters more directly.
1.2 BASIN MODELING AND STREAMFLOW FORECASTING
That aspect of hydrology known as Streamflow Forecasting undertakesto predict the outflow from a given catchment, in terms of flow rate as afunction of time, in response to a given precipitation event under giveninitial conditions. This capability is vital to effective planning for urban/industrial development, flood control hydroelectric power, navigation, andwater resources management.
Figure 1 depicts the cross section of a somewhat idealized rural.catchment and identifies the principal phenomena at work in the rainfall-runoff relationship. The input (precipitation) is partially intercepted byvegetation and water retention areas. Moisture reaching pervious surfacesdivides between overland flow, infiltration and evaporation. Through sub-surface processes, interflow and groundwater flow contribute ultimately tostreamflow, with some losses due to transpiration through plant life.
All the phenomena involved in the hydrological cycle are widely andwell understood qualitatively, and several empirical relationships have
1-1
PRECIPITATION
4 4, 4 EVAPORATION TRANSPIRATION
INTERCEPTION
RAINFALLGAGE
OVERLANDINFILTRATION FLOWAND PERCOLATION
.:.::::::::::..... .STREAM INTERFLOW
GROUNDWATERGROUNDWATER TABLE G DEEP FLOW
STORAGE
Figure 1. Cross Section of Idealized Rural Catchment
been developed. Integrating them into comprehensive parametric modelshas been achieved by several investigators, developments which werefacilitated by the availability of large, high-speed digital computers. Ofthese, the Stanford Watershed Model is probably the best known. In orderto apply any model to a given catchment, however, it is necessary that itbe calibrated. That is, several years of streamflow are simulated usingactual precipitation data, synthesized flows are then compared to actualrecorded flows, model parameters (as well as seasonal factors) adjustedby hydrologists and the process repeated until acceptable simulationaccuracy has been achieved. More recently, self-calibrating models havebeen developed which use trial and error routines to optimize model andseasonal parameters, but the calibration still required historical stream-flow data. If a stochastic model or technique is used instead of a parametricmodel, several dozen storm events in the form of previously measured rain-fall and runoff data must be used in multivariate statistical analyses in orderto establish confidence in the results.
Whatever the model or analytical techniques used, current methodsof streamflow forecasting require installation and maintenance of a systemof instruments so that data may be collected for several years. The basinof a major river must be divided into hundreds of watersheds of manageablesize and each one provided its own stream gage. Although the installationand maintenance expenses may not be objectionable, waiting for severalyears of data collection to initiate a new development may be intolerable.There is a real need for a method of predicting the hydrological behaviorof a previously ungaged catchment quickly (within two months, say, ratherthan three to five years) and with reasonable accuracy.
1.3 POTENTIAL APPLICATION OF REMOTE SENSING
There has been increasing interest in recent years in the relationshipbetween hydrological behavior of a watershed and a unique set of physio-graphic features by which it might be characterized. One would like to beable to quantify simulation model parameters, either directly or inferentially,from observations of catchment morphology and climate, taking advantage ofall available prior knowledge of the area. Earth observation spacecraft,manned or unmanned, make observations using a variety of sensors, cover-ing large areas in a short time. The objective of the study whose resultsare reported herein is to assess the extent to which the data provided byearth observation missions can be used to reduce the time, effort, andcost of predicting the hydrological behavior of a given catchment.
1-3
1.4 FEASIBILITY STUDY METHODOLOGY
The study was conducted in three phases as shown in
Figure 2, aimed at devising a set of widely applicable pre-diction model parameters in terms of photographically observable
characteristics (such as climatology, areas, elevations, and land use)and inferable characteristics (such as soil depth and porosity).
1.4.1 PRINCIPAL STUDY ELEMENTS
The principal elements of the study are a simulation and a cali-
bration model, a study area, and a complete historical and physiographicdata base pertaining to the study area. They are as follows:
* Simulation Model. The continuous simulation model chosen
for the study, for the reasons discussed in Section 2, isknown as the Kentucky Watershed Model (KWM). This model
is based upon the well-known Stanford Watershed Model IV,and has been further adapted and refined by IBM for application
to this project.
* Calibration Program. A calibration program known as OPSET(because it estimates the optimum set of watershed model
parameters) is a companion to the KWM. It has also been
refined and modified for application to this study. By a trial
and error process it estimates values for 13 simulation
model parameters which are further refined by manual
operations. This program is also described in Section 2.
* Study Area. The area from which test watersheds are chosenfor the study is the Tennessee River Valley, a major watershedof approximately 104,000 square kilometers in the southeasternUnited States. This is a well-instrumented, thoroughly photo-graphed and mapped area for which copious historical data recordsare available.
1-4
INSRU i MENTEDS:ATE SHEDI*T M HISTORICAL DATA)
IAGE IOCESSING FAINLLC
T EAMF. WITHOUT HISTORICAL DATA.
AND CLIMAT'OLOGICAL
ANALYSIS DA. TA
*DETERMINE WATERSHED PROCF SSINGPHYSICAL CHAACTERISTICS
LI-ATOLOGICAL
TVA WATERSHED DATA BASE CONTINUOUS
" HsroOcAL HYDROLOGICAL SIMULATION CORLATE MODEL CORRELATION MODE LS TREAMFLOW MODEL PAHTERS DERIVEMODEL PR TS CONTINUOUS SIMULATIONCLI ATOLOGICAL U CALCULATE MODEL THA MUWLAECOPA M IO
•ISTORICAL PHYSIOGRAPHIC PARAMETERS Y CAL aCORRELAF SIMULATE CONTINUOUS
MAPCOVERAGE * SENSITIVITY CHARACTERISTICS IATRSHEDPMYSICAL STREAMIFLOW
ARALPHOTOoCOVERAGE ANALYSIS a CHARACTERISTICS
I CALIBRATION CORRELATION 6 VALIDATION
Figure 2. Study Methodology
Within the Tennessee Valley 55 watersheds have been identified
as suitable to the purposes of the project, representing all six
physiographic provinces of the valley. It was estimated that 35watersheds would be needed for confidence in the results of thecorrelation process. Of these, 25 were to be used for calibrationand 10 for tests of simulation accuracy. Table 1 lists thelimited set actually used in the study. Figure 3 indicatestheir locations in the region. These represent five ofthe six physiographic provinces and range in area from 7 to 365square kilometers. The average overland flow surface slopesare indicative of the varying ruggedness of the region, fromthe Nashville Basin to the Blue Ridge Mountains. For all exceptthe smallest watersheds, hourly rainfall station records aresupplemented by daily rainfall records for the sake of accuracy.
Historical Data Base. The study requires accumulation of acomprehensive data base consisting of streamflow, precipitation,and evaporation records for the study area. The TennesseeValley watershed is one which is extensively instrumented andfor which ample historical data are available. It contains approxi-mately 560 rainfall stations, both hourly and daily. Almost 1,000streamflow records have been accumulated in the past severaldecades; approximately 360 streamflow stations are currentlyactive, and some 600 have been discontinued. Other climato-logical data such as temperature, evaporation, wind and humidityhave been carefully collected and catalogued.
Physiographic Data Base. The study also requires the accumulationof a comprehensive physiographic data base for the study area.Complete topographic map coverage is available, consistingalmost entirely of 7. 5 minute quadrangle maps derived fromaerial photographs at a scale of 1:24,000. The TennesseeValley has also been completely surveyed by aerial photography,and black and white photographs are available for every water-shed, mainly at a scale of 1:24,000. Some photographic coverageis also recently available in pseudocolor infrared taken fromhigh altitude aircraft, at a scale of 1:63,000. Excellent imagesfrom the Earth Resources Technology Satellite (ERTS) have alsobeen obtained.
1-6
Table 1. Selected Watersheds
PHYSIOGRAPHIC AREA AVERAGE NUMBER OF U*
CATCHMENT NAME (SQUARE RAINFALL STATIONS SPROVINCE (KILOMETERS) SLOPE Hourly Daily E
1. WHITE CREEK NEAR SHARPS CHAPEL, TN (WHITE HOLLOW) Valley and Ridge 6.94 .300 1 - C
2. LITTLE CHESTUEE CREEK BELOW WILSON STATION, TN Valley and Ridge 21.30 .237 1 1 C
3. SOUTH FORK MILLS RIVER AT PINK BEDS, NC Blue Ridge 25.70 .315 2 1 C
4. NOLAND CREEK NEAR BRYSON CITY, NC Blue Ridge 35.70 .500 2 - V
5. WEST FORK PIGEON RIVER ABOVE HAZELWOOD, NC Blue Ridge 37.30 .413 1 1 C
6. BIG BIGBY CREEK AT SANDY HOOK, TN Highland Rim 45.30 .185 1 1 V
7. BIG ROCK CREEK AT LEWISBURG, TN Nashville Basin 64.50 .155 1 - C
8. LITTLE RIVER ABOVE HIGH FALLS, NEAR CEDAR MTN., NC Blue Ridge 69.50 .246 1 2 V
9. TRACE CREEK NEAR DENVER, TN Highland Rim 78.70 .129 1 2 C
10. PINEY CREEK NEAR ATHENS, AL Highland Rim 145.00 .033 1 4 V
11. RUTHERFORD CREEK NEAR CARTERS CREEK, TN Nashville Basin 178.00 .133 1 2 U
12. EMORY RIVER NEAR WARTBURG, TN Cumberland Plateau 215.00 .298 1 3 C
13. SEWEE CREEK NEAR DECATUR, TN Valley and Ridge 303.00 .165 2 2 C
14. TOWN CREEK NEAR GERALDINE, AL Cumberland Plateau 365.00 .062 2 5 C
15. POPLAR CREEK NEAR OAK RIDGE, TN Cumberland Plateau 214.00 .329 1 3 C
16. EAST FORK POPLAR NEAR OAK RIDGE Cumberland Plateau 50.50 .215 1 1 V
*USE: C= CALIBRATION; V= VALIDATION; U= UNUSEABLE (INADEQUATE DATA)
SVA 37
KENTUCKY
WHITE HOLLOWSEMORY RIVER--- ,,--------- ,, I-,, VALLEYAND R/DGE
CUMBERLAND PLA TEA U (Including Sequatchie Valley)f(Including Lookout Mountain)
TENNESSEE
MISSISSIPPI 3.6 ALLUVIALN
C
PLAIN NASHVILLE BASIN N C AR
BL UE RIDGE
HIGHLAND RIM
' S CARLITTLE CHESTUEE
\ GEORGIA ""
MISS!ALABAMA
PHYSIOGRAPHIC PROVINCESOF THE
TENNESSEE RIVER BASIN
SCALE 0 It 20 30 40 50 MILES
FIGURE 3 LOCATIONS OF STUDY WATERSHEDS
1. 4.2 CALIBRATION
In the calibration phase the model parameters and their relatedobservables were quantified and adjusted for ten selected watershedsfrom five of the six physiographic provinces of the Tennessee Valley.These sets of satisfactory model parameter values are those whichyield simulated streamflows which best correlate with actual recordedflows for periods of five to ten years. A sensitivity analysis of twowatersheds provided guidance for final manual adjustments of the modelparameters, thereby minimizing the time and effort spent in the cali-bration phase.
1. 4.3 CORRELATION
In the correlation phase, satisfactory model parameters werecorrelated with observable physiographic characteristics in a multipleregression analysis, using the results of ten calibrations as a data base.The product of this phase is a set of regression equations, each of whichis an expression for a model parameter as a linear combination of otherparameters.
1.4.4 VALIDATION
In the validation phase, the relationships .produced by the correlationanalysis were tested on five watersheds from different physiographicprovinces of the region. No calibrations were performed on these basins;they were treated as "ungaged" watersheds. Their simulation modelparameters were quantified by using the regression equations, and thenthe simulation model was run, using those parameters and historicalclimatological data as input. The simulated streamflows compared wellwith observed streamflows, in respect to most indices of performanceindicating the feasibility of the technique.
1. 5 SUMMARY OF RESULTS
1. 5. 1 FEASIBILITY CLEARLY INDICATED
The five validation runs produced simulated streamflows whichcorrelated remarkably well with observed streamflow. Daily correlationcoefficients ranged from 0. 83 to 0. 87; monthly, from 0. 92 to 0. 97. Manymajor storms were reasonably well matched wi th respect to peak flowsand timing of peaks. For a multi-year open-loop simulation, this isadequate for most applications, and it strongly indicates the feasibilityof using remotely sensed data to forecast the hydrologic performance ofan ungaged watershed.
1-9
This happy conclusion should be kept in the perspective shaped bythe following facts and circumstances.
* The initial study plan envisioned using 25 watersheds forcalibration and ten for validation to provide statisticalconfidence in the results. The study actually used tenand five respectively, leaving some (undefined) un-certainty in the validity of the result.
* The remote sensing imagery used to determine watershedphysical characteristics was actually aerial photographs,mostly at a scale of 1:24, 000. No interpretation ofhigher-altitude (RB-57F, U-2) or space (ERTS) imageswas done in the study.
* Some ground truth was necessary, in this instance soil data.This is not considered a disqualifying fact, because the needfor some ground truth probably cannot be obviated in futureapplications.
* The correlation (i. e., multiple regression) analysis deservessome refinement, based on an enlarged appreciation by thestatistical analyst of the hydrologic processes and relationshipsfound in a watershed.
1. 5.2 SIGNIFICANT BY-PRODUCTS
There are two by-products of the study which the authors believeto be of potential value to other investigators in hydrologic modeling andrelated fields.
1. 5. 2. 1 Integrated Data Base
As in most applications using simulation models, the preponderantshare of the study effort was devoted to acquiring, verifying, formatting,and preprocessing input data to construct a data basis. As a result of thisstudy, a data base now exists in the form of a magnetic tape. The tape hasa capacity of up to 50 files, one file per basin. Each file has fields allocatedfor up to ten years of historical climatological data (mean basin hourlyrainfall, evaporation, temperature), mean daily streamflow, and selectedstorm events (up to ten per year, characterized by peak flow and time ofpeak). Each file also has fields for watershed features, simulation modelparameters and program control options.
At present, the data base contains data, except temperature, pertinentto the 15 Tennessee Valley basins used in the study. To any other investigatorinterested in doing similar studies in the same area, the integrated data basemay be of great value.
1-10
1. 5.2.2 The IBM System for Simulation and Analysis of Watersheds
Since the study team began adapting the Kentucky Watershed Modeland its companion calibration program, OPSET, to the feasibility study,several programming additions and modifications have been made. Theresult is a system of methods and software, built around the basic simu-lation and calibration models, which is a powerful hydrologic research toolreadily adaptable to operational use. Features of the system are as follows.
* Input data preprocessing and verification.
* Plot outputs.
* Tabular summaries.
* Storm analysis.
* Statistical analysis routines.
* Terminal operation for
set up and initiation of calibration/simulation runs,with a batch-processing computer, or
real-time interaction, with a time sharing computer.
The effort required to develop this system for efficient operation hasbeen more than compensated by the manpower saved in manipulation ofdata, handling many large decks of cards, and analyzing raw simulationoutputs. In fact, the study could not have been completed without it.
1. 6 A GLIMPSE OF THE FUTURE
Section 4 of this report summarizes research efforts worthy of con-sideration as sequels to this feasibility study. Several years will be requiredfor the idea tested preliminarily here to be reduced to practice in anotherregion. Nevertheless, one can readily visualize the sequence of events.
* A set of spacecraft sensors observe an ungaged region con-current with aircraft underflights and ground truth patrols.
* The data so acquired are analyzed and interpreted to yieldwatershed characteristics.
* Statistically derived relations are used to quantify simulationmodel parameters from observed characteristics.
1-11
* The simulation model is run, using climatological inputstypical of the region, to generate multiyear streamflowpredictions and statistics, for planning purposes.
This can be done in a short time, two to three months, without extensiveinstrumentation or the need to take calibration data for three or more years.Residual benefits will derive from implementing the system for periodicshort-term river forecasting, using a few well-placed stream gages for"closed-loop" operation and a forecasting accuracy within five per cent.
1-12
SECTION 2
SELECTION OF SIMULATION MODEL
2.1 OBJECTIVE
The spreading availability of high-performance computers has
stimulated remarkable progress in development of hydrological models.
An individual investigator may prefer one model or another on the basis
of several criteria, with emphasis on the objectives of his research. We
have selected a model with which we can accomplish the calibration,correlation, and validation phases within the study schedule time and
manpower budgets. The models and procedures developed must be
capable of predicting the hydrological performance of an ungaged basin
with parameters derived from space or aerial imagery, sparse ground
samples, precipitation and evaporation data.
2.2 STOCHASTIC MODELS
The project first considered stochastic models, which involved theuse of multivariate regression analysis to develop predictions of runoff
as a function of a limited number of observable variables: storm duration,storm intensity, time of the year, and initial moisture conditions--the lastestimated empirically from recent precipitation experience. Techniques
were identified for deriving rainfall-runoff relations using observable orinferable physiographic features. Synthetic hydrograph* methods wereconsidered for determining streamflow as a function of time. Methodswere identified for statistically correlating observable physiographicfeatures with unit hydrograph parameters.
2.3 PARAMETRIC MODELS
Parametric models were also considered and found to be preferableto stochastic models for the following reasons:
A hydrograph is simply a plot of streamflow in volume per unit time orriver height as a function of time. See Reference 2, Chapter 9.
2-1
1. The cause and effect relationships among the conditions andprocesses in a watershed are obscured in the stochastic modelbut are used explicitly in the design of the parametric model,many of whose parameters are observable directly from remotesensed imagery.
2. In using a stochastic rainfall-runoff (RF-RO) model, it isnecessary to analyze some 50 observations (storm events) perwatershed to obtain confidence in the validity of the derivedrelationship. In determining RF-RO relationships based uponphysiographic parameters, approximately 50 watersheds areneeded, each one having 50 associated storm events.
3. In order to take into account all practical variations in andinterrelations among watershed parameters, a prohibitiveamount of data is required for thorough statistical analysis.Thomas and Benson 3 identified 71 physiographic and statisticalvariables of interest in relating streamflow characteristics tobasin physical characteristics.
4. A parametric model that is capable of simulating continuousflows over a long period of time supports a larger variety ofplanning objectives than a method which is intended to estimateor predict a single variable or isolated event (e.g., flood peakor total runoff from a storm).
2.4 STANFORD WATERSHED MODEL (SWM)
The Stanford Watershed Model45 is probably the best known of theparametric hydrological models and, in all its modifications, is probablythe most widely used. Since it was originally published in 1962, severalreports have appeared in literature describing modified versions andapplications (References 5, 6, 7, 8, 9, 10, and others). As a proventool it was attractive to our project, whose scope does not embraceadvances in hydrological modeling.
The Stanford Watershed Model uses a moisture accounting systemto synthesize a continuous hydrograph from the following:
2-2
1. Recorded climatological data, precipitation, evaporation and
(for snowmelt situations) temperature,
2. Measurable watershed characteristics such as drainage area
and friction of the watershed in impervious surfaces, and
3. Parameters used in the computation process which are known
to vary in magnitude among watersheds but have not been
quantitatively tied to specific measurable watershed properties.
For example, one parameter indexes the capacity of the soil
of the watershed as a whole to retain water.
The third class of inputs requires a trial and error series of cali-
bration runs to quantify a set of model parameters which will synthesize
flows with acceptable accuracy.
Figure 4 depicts the accounting of moisture entering the watershed
until it leaves by streamflow, evapotranspiration, or subsurface outflow.
A series of relations each based on empirical observation or theoretical
description of a specific hydrologic process, is used to estimate rates
and volumes of moisture movement from one storage category to another,
in accordance with current storage states and the calibrated watershed
parameters. The model routes channel inflow from the point where it
enters a tributary channel to the downstream point for which a hydro-
graph is required. A subroutine exists for including snow in the accounting
but is not needed in the present study.
2.5 KENTUCKY WATERSHED MODEL (KWM) AND OPSET PROGRAM
The Stanford Watershed Model was originally written in the Burroughs
Computer Language (BALGOL) then in use at the Stanford Computer Center.
It has subsequently been translated into Fortran IV, and a number of
adaptations were introduced in one version to suit the climate and
geography of Kentucky as representative of the eastern United States.
In a recent research program a version of the model using an initial
set of model parameter values and a number of control options was
developed for use with a self-calibrating streamlined version of the
model. These models are referred to as the Kentucky Watershed Model
(KWM) and OPSET (because it estimates the OPtimum SET of model
2-3
PRECIPITATION OR SNOWPACK RUNOFFFREE WATER POTENTIAL EVAPORATION
EVAPOTRANSPI RATION
ACTUAL IMPREVIOUSACTUALSET I
4 -- EVAPO. _ INTERCEPTION INTERCEPTIONTRANSPIRATIONT STORAGEIERCEPTN
SURFACERUNOFF
INFILTRATION UPPER ZONERETENTION
INTERFLOW
TRANSPIRATION RETENTION
TRANSPI RATIONS
ACTIVE OR DEEPGROUNDWATER
- - EVAPOTRANSPIRATION l- - - - - -- --- _ ACTIVE GROUNDWATERSTORAGE
DEEPORINACTIVE / INFLOW TO \GROUNDWATER ( CHANNELSTORAGE \ SYSTEM
LEGEND
OUTPUT
Figure 4. Moisture Accounting in the Stanford Watershed Model
2-4
parameters). The availability and utility of these models and reports
describing them 11,12, 13 led to their use in the IBM project.
2.5.1 KENTUCKY WATERSHED MODEL
Figure 5 lists the inputs (exclusive of control options) used by the
Kentucky Watershed Model to simulate streamflow. Climatological data
can be obtained from precipitation records or can be hypothetical, the
latter being useful in generating rainfall-runoff predictions. The inputs
classed as "Overland Flow Parameters" and "Watershed Parameters"
are readily obtainable from analysis and interpretation of images (maps
and/or photographs). The inputs on the right side of the figure are
estimated in the calibration phase by OPSET. Some additional manual
calibration is necessary to develop a set of model parameters that best
represents the watershed.
2.5.2 OPSET
When a user applies a simulation model to a watershed, there are
several parameters whose values he must initially guess and subsequently
adjust, between trial runs of the model and comparisons of synthesized
with observed flows. This trial and error calibration requires ingenuity,familiarity with how the parameters interact within the model, and someunderstanding of the sensitivity of simulated flows to specific parameteradjustments. The process is aided greatly by a thorough understandingof the hydrologic process and by the guidance published by Crawford andLinsley 4 , 5. Through careful parameter adjustment, one can cause sim-
ulated flows to approximate recording flows but never to match them
exactly. Several combinations of parameter values can produce com-parable results from an overall viewpoint, and the final choice may well
hinge on whether a particular comparison emphasizes flood peaks, annual
runoff volume, or some other hydrograph feature. The final acceptanceof a set of parameters may depend heavily on subjective factors.
In developing OPSET, Liou 1 2 provided a tool for calibrating the
KWM with a minimum of subjective decisions. The parameter optimi-zation concept is depicted in flow chart form in Figure 6. The input dataconsists of control options and initial conditions as well as the inputslisted in Figure 7. A simulation is performed, one year at a time, using
2-5
OVERLAND FLOW MANNING'S "N" IMPERVIOUS INTERFLOW RECESSION CONSTANT IFRCSURFACES (M/P) OFMNIS 1 RECESSION
OVERLAND BASE FLOW RECESSION CONSTANT BFRC CONSTANTSFLOW OVERLAND FLOW MANNING'S "N" (M/P) OFMN -
PARAMETERS I LOWER ZONE STORAGE CAPACITY LZCOVERLAND FLOW SURFACE LENGTH (M/P) OFSL
BASIC MAXIMUM INFILTRATION RATEOVERLAND FLOW SURFACE SLOPE (MIP) OFSS WITHIN WATERSHED BMIR
MONTHLY PAN COEFFICIENTS SEASONAL UPPER ZONE STORAGESCAPACITY FACTOR SUZC LAND PHASE
MEAN ANNUAL NUMBER OF RAINY DAYS MNRD PARAMETERS
CLIMATO- EVAPOTRANSPIRATION LOSS FACTOR ETLFLOGICAL DATA ESTIMATED POTENTIAL ANNUAL EVAPORATION BASIC UPPER ZONE STORAGE CAPACITYBASIC UPPER ZONE STORAGE CAPACITY
EPAET < FACTOR BUZCOBSERVED RAINFALL DATA SEASONAL INFILTRATION ADJUSTMENT
O CONSTANT SIAC
.j SIMULATED STREAMFLOWUJ
OE INTERFLOW
LAND USE (TYPE & DENSITY) (M/P) VINTMR Z BASIC INTERFLOW VOLUME FACTOR BIVFNo VOLUME
PARAMETERFRACTION OF WATERSHED COVERED BY WATERSURFACES (M/P) FWTR CHANNEL STORAGE ROUTING INDEX CSRX
WATERSHEDWATERSHED CHANNEL
PARAMETERS FRACTION OF WATERSHED COVERED BY _ FLOOD PLAIN STORAGE ROUTING INDEX FSRX ROUTINGIMPERVIOUS SURFACES (M/P) FIMP U) CHANNEL CAPACITY-INDEXED TO BASIN PARAMETERS
OUTLET CHCAP
AREA OF WATERSHED (M/P) AREA
TIME AREA HISTOGRAM OF WATERSHED(M/P) NBTRI, BTRI
NOTE: M/P - MAPS AND/OR PHOTOGRAPHS
Figure 5. Simulation Model (KWM) Inputs and Output
DATATAPE
INPUT DATA . OBSERVED FLOW DATA
IMULO SYNTHESIZED FLOWS I COMPARISONSIMULATIONOF STATISTIC
PARAMETER NO E
MATCH
ADJUSTMENTMADE?
YES
OPTIMUMPARAMETER
VALUES
Figure 6. Parameter Optimization Concept
OVERLAND FLOW MANNING'S "N" IMPERVIOUS INTERFLOW RECESSION CONSTANT IFRCSURFACES (M/P) OFMNIS 1RECESSION
BASEFLOW RECESSION CONSTANT BFRC CONSTANTSOVERLAND OVERLAND FLOW MANNING'S "N" (M/P) OFMNFLOWPARAMETERS LOWER ZONE STORAGE CAPACITY LZC
OVERLAND FLOW SURFACE LENGTH (M/P) OFSL
BASIC MAXIMUM INFILTRATION RATEOVERLAND FLOW SURFACE SLOPE (M/P) OFSS WITHIN WATERSHED BMIR
MONTHLY PAN COEFFICIENTS SEASONAL UPPER ZONE STORAGE LAND PHASECAPACITY FACTOR SUZC PARAMETERS
MEAN ANNUAL NUMBER OF RAINY DAYS MNRD
CLIMATO- - EVAPOTRANSPIRATION LOSS FACTOR ETLF
LOGICAL DATA ESTIMATED POTENTIAL ANNUAL EVAPORATION LU BASIC UPPER ZONE STORAGE CAPACITYEPAET k FACTOR BUZC
OBSERVED RAINFALL DATA O SEASONAL INFILTRATION ADJUSTMENT- CONSTANT SIAC
OBERVED STREAMFLOW DATA SIMULATED STREAMFLOW
0 P,
INTERFLOWLAND USE (TYPE & DENSITY) (M/P) VINTMR 1 - BASIC INTERFLOW VOLUME FACTOR BIVF INTVOLUME
z PARAMETER
FRACTION OF WATERSHED COVERED BY WATER O NUMBER OF CURRENT TIME ROUTINGS INCREMENTS NCTRI
SURFACES (M/P) FWTRWATERSHED
TERSCHANNEL STORAGE ROUTING INDEX CSRXPARAMETERS FRACTION OF WATERSHED COVERED BY a CHANNELIMPERVIOUS SURFACES (M/P) FIMP ROUTING
IMPERVIOUS SURFACES MP FMP FLOOD PLAIN STORAGE ROUTING INDEX FSRX PARAMETERS
AREA OF WATERSHED IM/P) AREA CHANNEL CAPACITY-INDEXED TO BASINOUTLET CHCAP
TIME AREA HISTOGRAM OF WATERSHED(M/P) NBTRI, BTRI
NOTE: M/P - MAPS AND/OR PHOTOGRAPHS
Figure 7. Calibration Program (OPSET) Inputs and Outputs
a "streamlined" KWM. The synthesized flows are compared with the
observed flows. An objective function is used to determine when an
optimum set of parameters has been found. If the best match has not
been achieved, parameters are again adjusted and the simulation run
again. This sequence is repeated until a satisfactory parameter set has
been quantified.
Figure 7 also lists the 13 outputs of OPSET, in addition to simulated
streamflow. (Comparison with Figure 5 shows the relationship to KWM.)
These parameters are the most difficult to measure directly and ones to
which simulated flow values are sensitive. Yet they, too, must be esti-
mated from observation and correlation in the future, to apply the
simulation model to an ungaged watershed. The calibration process
should be based on three separate water years for the same basin.
Simulation model parameters are then derived by averaging the results
of the three calibration runs. A minor modification to OPSET has been
implemented to generate a more precise Base Flow Recession Constant
(BFRC). As it is presently designed, OPSET estimates parameterswhich produce accurate simulations of major winter storms (withrespect to flood peak magnitude and timing) but misses summer and
autumn storm peaks by significant factors. Manual adjustments arerequired to achieve accurate simulation in the latter. An improvement
in OPSET efficiency could be achieved by modifying it to calibrate on
the basis of several consecutive years rather than one year at a time.
2.6 MODIFICATIONS AND ADDITIONS TO KWM AND OPSET
In adapting the KWM and OPSET programs to the project, severaladditions and modifications were implemented by IBM for improvedflexibility and efficiency, both in running the program and evaluatingresults. They are listed as follows:
1. The subroutine called READ was modified to operate on anycurrent System/360 installation.
2. Both models are in the Production Library, and the input datais in a partition data set (PDS) based operation which will (a)avoid the need for maintaining several thousand cards for eachwatershed, and (b) permit operation from a terminal, giving theinvestigator quick access to the data base for modifications.
2-9
3. A subroutine designated INT has been added to convert
historical stream gage height information from strip charts
to streamflow rates (volume per unit time), applying the
gage rating tables. It also analyzes the peaks and calculates
the runoff.
4. A subroutine designated STAT has been added to provide
statistical analysis of monthly and daily data for a single
year, and multiple years. Sample output for a yearlystatistical summary is shown in Figure 8, and the total
statistical summary output for seven years of simulation
is presented in Figure 9.
5. An Output Summary routine tabulates all the end results of
a calibration or simulation run for a particular watershed
so that the investigator need not search through several
pages of printout to compare and evaluate results. Figure 10
presents an Output Summary.
6. As it was originally received, the KWM could performsimulations only for one year at a time, and it was necessaryto re-establish initial conditions as well as re-read watershed
descriptive data before each run. The capability has been
introduced to simulate any number of successive years with
only one reading of the watershed description and retaining
simulated moisture conditions at the end of each year to
serve as initial conditions for the succeeding year, thereby
providing more accurate simulation results.
7. A Plot Output routine with several options was introduced inorder to minimize manpower and time needed for evaluationof results, sensitivity analysis and calibration "fine tuning. "Plots shown in a subsequent section of this report were pro-duced by this program. Vertical (flow rate) and horizontal(time) scales may be specified or automatically chosen. Plotsare available for simulated flow, observed flow or the two
superimposed on the same scale. Time intervals availableare one year (mean daily streamflows), one month (mean dailystreamflows) and storm duration (hourly streamflows for theduration of the selected storm runoff).
2-10
*LITTLE CHESTUEE.BELOW MILSON.TENN. (8.24 S0.MI.) MYSI STUDY LCoB
AYFARI Y TATI~TfCAI IMMARY
MONTHLY DAILY
OBSERVED SIMULATED OBSERVED SIMULATED
_EAN 476,05 qA anRA 14 6i 1 ..__
MAX IMU 1101.00 11~6A07 ARAR-00 4n-R4
_-YARI ANCE 123591.00 9099QR56 7R80.4 R9 8n
IANDARDJ FDEVIATtoN 951 7 -nQ AA 796 47 .75
SUM OF (OBSERVED - SIMULATED) 542.01 541.97
ROOT SUn SQUARE 329.05 192.07
SUM SQUARED 1.34 134.05
SUM SQUARED (IBM METHOD) 1.36 136.00
CORRELATION COEFFICIENT 0.9758 0.9553
Figure 8. Example of Yearly Statistical Summary
*LITTLF CH STUEEt!3ELOW WILSON,TENN.(8.24 SQ. MI.) WY56 STUDY LC"'
TOTAI STATISTICA SUIMMARY
MONTHLY DAILY
OBSERVED SIMULATED OBSERVED SIMULATED
"FAN 395.90 424.25 .13.n3 13.94
X I MUM 163.O00 1458.49 388.00 507.47
VA IA NF 114971.44 12 11R.t sO 6 7_7- Z L717Ql
i- STANnPO FVIAT N I139.7 I.-R 24. 69i 32.74
SUM OF (OBSERVED- SIMULATED) -2381.21 -2381.01
ROOT SUM SQUARE 957.08 751.03
SUM SQUAR ED 12.58 1169.41
SUM SQUARED (IRM METHOD) 12.59 1192.33
CORRELATION COEFFICIENT 0.9576 0.9044
Figure 9. Example of Multi-Year Statistical Summary
LITTLE CHESTUEEJBELOW WILSON,TENN. (8,24 SQ.MI.) WY5L STUDY LC08
ICT N3V DEC JAN FEB MAR APR MAY JUNE JULY AUG SEPT ANNUAL
PRECIPITATION 3.300 2.210 3.210 4.440 4.960 9.330 4.200 1.850 4.870 3.970 2.270 6.520 51.130 IN;HESFVP/TRA-JNET 2.409 1.510 0.881 0.7)3 0.727 1.O~A 2.370 1.981 4.757 4.774 3.48 1.9Q iO.3R6 INPFZ
-POTENTIAL 2.068 1.568 0.887 0.723 0.727 1.044 2.448 4.658 5.975 6.638 6.864 5.028 -9.530 INCHESSURFACE RUNOFF 0050 0.060 O.141 0.791 2.086 3.1Al 1.011 0.064 0.34& 0.11& .011 0.6 R8.301 INHFINTERFLOW 0.0 0.0 0.0 0.025. 0.188 U.599 0.299 0.010 0.001 0.0 0.0 U.U 1.121 IN;HES_RASE F1LW 0.728 0.593 0.642 0.680 0.846 I.26 1.543 1. 411 1.20~7 0.RI 0.31 11.20 I1NfHEr -
STREAM EVAP. 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 IN:HESTOTAL RUNfFF(SIMI 0.778 0.653 0.783 1.497 1.120 5.218 2).82 1.10 1.462 1.190 0.842 0.876 lO .a2 IN'M HTOTAL RUNOFF(REF) 0.947 0.813 1.617 1.867 3.643 5.872 3.429 1.212 0.980 1.053 0.735 0.908 23.075 IN;HES
IBSERVED TOTALS 209.8 180.2 358.3 413.6 807.1 1301.0 759.7 268.6 217.1 233.2 162.8 201.2 5112.6 CFSSIMULATFr TnTALA 172.4 144.6 173.6 111.7 491.2 1156.1 631.9 301.3 13.9 263. 186.5 14.0 4570.6 CFP
BALANCE -0.0229 INCHES
MOTHLY FLOW CORRELATION COEFFICIENT 0.9758MFAN DAILY FIwM IORRFIATION COFFFICIFNT 0.9853
Figure 10. Example of Output Summary
The subroutine INT, added to KWM for storm analysis, constitutesa major modification. It has the following capabilities:
* Converts gage height (ft) to streamflow (cfs), a task whichformerly had to be done manually.
* Calculates the total runoff volume resulting from each storm.
* Summarizes the characteristics of each selected storm: peakflow rate, time of peak, etc.
Evaluation and analysis of hydrograph characteristics requiresobserved hourly streamflow data. An example of available observed*hydrograph data is shown in Figure 11. The observed hydrograph datais in terms of stream gage height (ft), but the simulated hydrograph datais in streamflow (cfs). Therefore the observed data must be convertedto cfs, which requires rating tables for for that period of time. A samplerating table is shown in Figure 12. This rating table is valid for the periodfrom February 11, 1953, to March 9, 1953. The present study uses seven yearsof historical data (1950-1956) and requires 17 rating tables. The subroutineINT uses the proper rating table to convert stream height to flow rate.Operating with selected major storms, rather than a continuous 8,760 hoursof data per year, reduced the effort associated with observed input data(manual translation of gage height hydrographs) and processing time by afactor of 40, assuming an average storm duration of 48 hours and fivemajor storms per year. The subroutine INT has the capability of integrationto calculate the runoff volume produced by observed and simulated storms.
The calibration and simulation models and their data base have beenadapted to operation from data terminals in the IBM-Huntsville facility.Access to an IBM System/360 Model 75 is afforded from a Model 2260terminal, from which an operator can call up sections of the data base
*Throughout this report, the term "observed" refers to recorded historicaldata, distinguished from the "simulated" or "synthesized" results producedby the simulation model.
2-14
J 7.
A. . 1 . j.. . , 1-IA
.. ... ....,- .. ... ....... .. . .. .... .-. ., I s , .- • ..,.
L J
Figure 11. Example of Observed Hydrograph DataI Reproduced frombest available copy.
Tennessee Valley Authority9-210 Hydraulic Data Branch ashigon.......................
(Sept. 1952) Fle No. ....
Rating tabic for.. e...._.C..... L .. C / ...A t. .... . . ... . .........................
......................... ......................... ....... , from . L./ , ...... 9 3..., tlo ... ..l ... .. ..... 9.
S c rge Dieff..- . ]) tI- I tC.' )"c- ' I ic h o (: " .i:r -"
Cigh t Lc ro C. ,t 1) e hei:ht C wfrs hc :ijt
I l
r -E -
FC , cy Cs u ' Cs c(I Ft C/s Cfs Ftd C/s Ft, cs C
. ----- 6-- - .0 -..- .0 . ...... ..
. ... .. . .......... .... . 0 . .
.20 ---------------------- 30. a------------------------ ---- --------------------------------------------- ---------
.o ............ ..... - " . -. o . ............. . .o
w -- .--
. ............. ..... . .. . .............. .20 ...........
.o ... . . ...... ...... ........... ... 0 ----...............-------------- ... oS... ... . .... .... .............. o ...... . ..
.-- ..--. ... -. -
Ao ~~~~~------- ----i.---------------------------
. . ..... .............. .7 . ..........._ . . 0 .... ... . .. .
. ..... 7.... . ... -_. . .o . ...... . .o ....j ..0. .- . .........
0 .90 .60------------
---------------------------_________________ _______-----_______0 -------------- .0--------------
he above table is not applcabl_ for ice or obstructed channel condit.o.s. It s based on--.-- .....discharge measure-
Smen~. mde durig .. l . ----------..................
and ..... ~..------ ... . .o - .............------------- nd ..............-------------.-- 13 - - - *-.. ... ........
............ .------. . .... . .-------- ---------- ------ ---- ----..-............ 2 ... ........
. - - --....-,-....... ............. .o . . ...... . .............- -c - --- -- . -.. ............
------- -- - I- .30
------- ----------------------- .. -------------------------............... . ---- ........... . ..............
--------- ---------- ------- - ------------- ~
.0 --- --- Z J ....... ...... . .. oI so1 .90 -- ---- -- s ---- --- .60 -------
--------- - ------------------ -------- Dt --------.
3, ~3 1------------ ----------- :: _
ThFigure 12. Typabl ical Sot pplicble or icage or otrucingd channel conditions. et 's hased on ---..... discharge measure-M een t m cid e d u rin g ...------------- .-------------------------------------------- ------------------------- .------------------------------------------------------
----- -------- -- ---------- ------------ ------------------ ------ ------------- ------ .~~..~-- ~ ~ ---~~---------- --- ---
and is .......................................................... well defined bc -.: ............... .........----- cfs and ...-................... .cfs.
--- ~--~~-~ ~ -~ ~~-~~~-'- -- ~~~~ -~ ~ "~~~ ~-~ ~ ~-"-----------------" ~ ~~
.----.----------------------- --- -- -- ---.- --........ ...-..-.-... . . . . . . . . .......... -- - -... -- .. . . . .. . . . .. .. ....-... ,-.......................
-------------------------------------------------------------------------------- .~. .. ~.~..~~..~.. Compted by -. ASC F -----------------~
--- ------------ ----- --------------------------------------------- --------------------------------..- Checkcd bj ------------------------------
-. .-.------------------------ .. ..-.-..-.-.-.----------.--.----------------------------------------------------....--.Date................ .......
v. s. rOvLahkt.iL PkLItiNG Orrc If,, niq t' 3
Figure 12. Typical Stream Gage Rating Table
2-16
and modify elements thereof, enter control options in the program,
change model parameters, and enter the program in the computer
system queue, without having to manipulate punched cards. In the
sensitivity analysis (explained in a subsequent section of this report),
several runs can be set up and initiated at one setting. It is conserva-
tively estimated that this terminal operation approach has obviated
the storage and maintenance of 500,000 punched cards and improved
the efficiency of operation by a factor of ten.
2-17
SECTION 3
TECHNICAL RESULTS
This section describes the principal and secondary accomplishmentsof the feasibility study. The study consists of nine tasks, identified asshown in Figure 13, which illustrates the flow of study activities. In theremaining pages of this section, each task is described with respect toits objective, the technical approach used, and the results achieved.
Physiographic data were collected (Task 1) in the form of topographicmaps, aerial photographs, and other documents for 15 watersheds of theTennessee Valley. The historical data (Task 2) consists of observedstreamflow, precipitation, and evaporation data related to the same 15watersheds. All these physiographic and historical data are digitizedand integrated into the Master Data Bank (Task 3). (Note: Although only15 basins were used in the study, the study team's accumulation of rawdata pertains to some 50 basins.)
Ten of the test watersheds were modeled, and the models werecalibrated using OPSET and manual adjustments (Task 4) in order toderive a set of optimum model parameters for each watershed. Thisprocess was aided by the sensitivity analysis (Task 5), which was per-formed on two of the selected watersheds. The sensitivity analysisshows quantitatively how the accuracy of the simulation model is affectedby variations in the model parameters.
The physiographic data base yields watershed characteristics foreach of the 10 basins on which a calibration will have been performed.These watershed characteristics, which are the observable or the in-ferable characteristics determined from remote sensing, were usedwith the optimum model parameters to perform a correlation analysis(Task 6) and thereby determine statistical relationships between optimummodel parameters and observable physical characteristics of the watershed.
Five watersheds were reserved and used to test the relationshipsdeveloped in the calibration and correlation tasks. Observable charac-teristics were used, together with the relationships determined by Task6, to generate sets of simulation model parameters. For each of thesewatersheds, the simulation model was then run, using actual climatolog-ical data as input. The simulated streamflow was then compared to theactual historical streamflow, to yield an evaluation or validation (Task 7)of the relationships.
3-1
TASK 2 FINAL
REPORT
HISTORICALDATA BASE
TASK 3 TASK 4 TASK 5 ASii TASK 9
MASTER CALIBRATION SENSITIVITY R ESOLUTION COMPREHENSIVE
DATA BANK ANALYSIS ::Ciiiii:iii : 'a REPORT
ANLISANALYSIS"REPORT(15 BASINS) (10 BASINS) (2 BASINS)
TASK 1 TASK 6 TASK 7WATERSHED
PHYSIOGRAPHIC CHARACTERISTICS CORRELATION PARAMETER - PHYSIOGRAPHY . VALIDATION
DATA BASE (15 BASINS) RELATIONSHIPS
(5 BASINS)
WATERSHED CHARACTERISTICS (5 BASINS)
HISTORICAL DATA AND WATERSHED DESCRIPTIONS (5 BASINS)
Figure 13. Feasibility Study Flow
Task 8, Resolution/Accuracy Analysis, was to be a determinationof the performance characteristics of remote sensors needed to makeapplication of the techniques practicable. The study was not able to ad-dress this task within the originally planned budget and schedule, but avery similar task is to be performed under another contract.
This report is the result of Task 9.
3-3
3.1 PHYSIOGRAPHIC DATA BASE
Operation of the mathematical models (both OPSET and KWM) usedin the hydrological simulation activity is governed by certain input para-meters which are related to the physical nature of the watersheds beingsimulated. These parameters must be measured or in some instancesinferred from existing documentation such as topographical maps, aerialphotographs, soil catalogs, survey reports, and other a priori information.This collected documentation, together with certain measured or inferredinformation therefrom, constitutes the physiographical data base whoseconstruction is the objective of Task I. Figure 14 depicts the activitiesinvolved in establishing the physiographic data base. Table 2 is a summaryof the applicable documentation presently available in IBM files, and Table3 is a physiographic data base status summary for the 15 watersheds utilizedin the feasibility study.
The physiographic data required by the mathematical models is aseries of parameters as listed and defined in Table 4. The methodologyemployed to derive these parameters from the physiographic documenta-tion is discussed in the following paragraphs.
3.1.1 AREA
The first step in determining the area of any given watershed is tolocate on a topographical map or photomosaic the streamflow gage asso-ciated with that particular watershed. Then beginning at the stream gagelocation, an outline of the watershed is manually drawn by following con-tour lines and ridges until the area perimeter closes on itself back atthe stream gage. Visualization of ridges, slopes, and drainage patternsis considerably enhanced by accentuating the waterways associated witha particular watershed. Accentuation is easily accomplished by darken-ing the rivers and streams on the map with a dark blue pencil. Once thewatershed perimeter is determined, the area can be measured with aplanimeter.
3.1.2 OVERLAND FLOW SURFACE LENGTH (OFSL)
The overland flow surface length is defined as the average distanceexcess precipitation, or runoff, must travel from ground impact until itreaches a permanent water course such as a river, stream, creek, branch,or crevasse. The average value of OFSL is determined by selecting 10 to20 points at random throughout the watershed as outlined on the topograph-ical map. A line perpendicular to the contour lines is then drawn from the
3-4
MAPFILES
MAPS
PHOTOMOSAICS
Overlay Graphics* INTERPRETATION
AERIAL " * ANALYSIS .PHOTOGRAPHS
* MEASUREMENTWatershed
Characteristics
SOIL CATALOGS * CALCULATIONcn1
* CLASSIFICATION
SURVEY REPORTS -
Simulation ModelParameters
OTHER A PRIORIINFORMATION
TO DATA BANK, MdTASK 3 Parameters
Figure 14. Physiographic Data Base
Table 2. "Raw" Physiographic Data Acquired(Tennessee Valley Region)
146 TOPOGRAPHIC MAPS, 7.5' QUADRANGLE
385 AERIAL PHOTOGRAPHS, USDA B&W; 1:15,000, 1:24,000, and 1:36,000
45 AERIAL PHOTOGRAPHS, NASA RB-57F, B&W PRINTS FROM IR NEGATIVES,1:120,000
250 GEOLOGIC AND MINERAL RESOURCE SUMMARIES
MISCELLANEOUS SPECIAL REPORTS AND DOCUMENTS
O
Table 3. Physiographic Data Base Status
MAP PHOTO PARAMETERSWATERSHED COVERAGE COVERAGE QUANTIFIED* REMARKS
WHITE HOLLOW A NONE 12/12 PHOTOS REQUIRED TO
LITTLE CHESTUEE PARTIAL 12/12 ESTIMATE VALUES FOR:
S. FORK MILLS RIVER PARTIAL 7/12 (1) FIMP
NOLAND CREEK PARTIAL 7/12 (2) VINTMR
W. FORK PIGEON RIVER PARTIAL 7/12 (3) OFMN
BIG BIGBY PARTIAL 7/12 (4) OFMNIS
BIG ROCK 100% PARTIAL 7/12
LITTLE RIVER PARTIAL 7/12 ALL OTHER PHYSICAL
TRACE CREEK NONE 7/12 PARAMETERS (EXCEPT
PINEY CREEK NONE 7/12 CHANNEL CAPACITY-
RUTHERFORD CREEK NONE 7/12 CHCAP) CAN BE
EMORY RIVER NONE 12/12 DETERMINED OR
SEWEE CREEK NONE 7/12 INFERRED FROM
TOWN CREEK NONE 7/12 TOPOGRAPHIC MAPS.
*Expressed as a fraction of the total number of parameters to be quantified.
Table 4. Physiographic Data Parameters
AREA Area of the Watershed in Square Miles
OFSL Average Overland Flow Surface Length in Feet
OFSS Average Overland Flow Surface Slope
TAH Time Area Histogram Parameters
FIMP Fraction of Total Watershed Area Covered by Impervious Surface
FWTR Fraction of Total Watershed Area Covered by Water Surface
VINTMR Vegetative Interception Maximum Rate in Inches/Hour
OFMN Overland Flow Manning's "N" for Pervious Surfaces
OFMNIS Overland Flow Manning's "N" for Impervious Surfaces
GWETF Ground Water Evapotranspiration Factor
SUBWF Subsurface Water Factor
(The significance and methods of quantification of these parameters are described in the text.)
selected point to the nearest water course. The length of that line is thenmeasured and tabulated. This same procedure is repeated for all selectedpoints and the average distance of the 10 to 20 points is calculated and re-corded as OFSL.
3.1.3 OVERLAND FLOW SURFACE SLOPE (OFSS)
The OFSS is the average slope of the land contained within theboundaries of the watershed and is determined as follows. For con-venience, the same points selected for OFSL may be used to determineOFSS. From each point selected, the height differential from that pointto the nearest watercourse can be determined by counting the contourlines and multiplying by the contour scale. Since the length of the over-land flow surface at that point was measured when determining OFSL,OFSS is then equal to the change in height (tz h) divided by the averageof the local OFSS values.
Table 5 is a sample worksheet showing the final average OFSS andOFSL values for a typical watershed.
3.1.4 TIME AREA HISTOGRAM
A time area histogram requires the division of the watershed into"n" subareas where each area is defined as a function of drainage time.Boundaries of subareas are known as isochrones, which may represent15-minute or one-hour intervals depending on the size of the watershed.A minimum of four isochrones is desired for operation in the simulationmodel. The steps required to develop a time area histogram are asfollows:
1. Utilizing the topographical map of the watershed,determine the longest watercourse in units of feetfrom the stream gage location to the furthermostpoint of the defined watershed. This will be notedas a stream length (L).
3-9
Table 5. Worksheet for OFSS/OFSL (Typical)
Site Overland Altitude LocalSelected Distance Change Slope
(No.) (Ft.) (Ft.) (AH/S)
1 1,000 160 0.1602 900 100 0.1103 1,400 260 0.1864 1,200 120 0.1005 2,000 120 0.0606 1,100 80 0.0737 1,500 160 0.1078 700 - 160 0.2299 2,000 180 0.090
10 700 180 0.258
12,500 1.333
OFSL = 12,500 - 10 = 1,250 ft.
OFSS = 1.333 - 10 = 0.1333
NOTE: Above values are actuals for Rutherford Creek near Carters Creek, Tennessee
2. Determine the change in elevation in feet between the stream
gage location and the furthermost point in the basin as
established in Step 1. This change will be noted as Ah.
3. Calculate the concentration time (Tc) in minutes utilizingthe empirical equation14
Tc = 0.0078( .77
where L is stream length and Ah is altitude differenceas defined previously. Concentration time is the time ittakes for a drop of water to reach the gaging station fromthe watershed extreme perimeter. If L and Ah are measuredin meters rather than feet, then the coefficient should be0.0195 rather than 0.0078. When Tc is equal to or greaterthan 240 minutes, the isochrone intervals should be one hour.This will provide the minimum subareas required to operatein the simulation models. When Tc is less than 240 minutes,then the isochrone interval should be 15 minutes. Since it isvery unlikely that Tc will calculate in exact multiples of 15minutes, the concentration time should be arbitrarily adjustedto the nearest number divisible by 15.
4. Calculate the number of subareas or isochrones by dividingthe concentration time by either 15 minutes or 60 minutes.
5. Calculate the average stream velocity (V) by dividing thestream length (L) by the concentration time (Tc).
LTc
6. Calculate the average overland flow distance per isochrone bymultiplying V by either 15 minutes or 60 minutes.
3-11
NOTE: Figure 15 shows an example of the calculations used to establish
a time area histogram.
Once the average overland flow distance per isochrone has been
determined, the actual subdivision of the subdivision of the watershed
can be accomplished as follows:
1. Beginning at the stream gage location on the topographicalmap and using a map measuring meter, track several water-courses from the gage station inland to a distance equal tothe average overland flow length per isochrone for thatparticular watershed. Measured distances must alwaysfollow the watercourses. Several measurements should bemade radially from the stream gage and a mark placed ateach overland flow distance milestone. A line connectingthese marks should then be drawn from perimeter toperimeter and this area identified as Isochrone 1. Thisisochrone identifies that area of the basin which would drainwithin the first 15 minutes (or one hour) after a storm. Thenmeasuring from the first isochrone line, repeat the processto establish the second isochrone area. This is the area thatwould be drained from 15 to 30 minutes after a rain. Continuethe process until the entire basin is subdivided into "n" areas.Figure 16 is an illustration of a typical watershed so divided.
2. Once the watershed is subdivided, measure with a planimeterthe area of each isochrone. These areas should be tabulatedas both absolute magnitude in square miles and also as afractional portion of the entire watershed. Table 6 is a typicaldata sheet with this information. Each isochrone area andfractional portion value is converted to punch cards andentered into the simulation model.
3.1.5 IMPERVIOUS SURFACE FRACTION (FIMP)
This is a fraction of the total watershed area covered by pavings,rooftops, streets, etc., whose runoff contributes directly to a stream.This value is usually estimated from aerial photography. The value isusually zero for rural areas except where there are large areas ofexposed rock adjacent to watercourses.
3-12
Maximum Stream Length (measured): 110,000 Ft.
Maximum Altitude Drop - H(Map Scale): 480 Ft.
0.77 0.77
L 110,000Tc = 0.0078 = 0.0078
AH 480
L 110,000
- 482 Minutes * 480 Minutes
Nisochrones = 480 - 60 = 8
V = 110,000 Ft./480 Minutes = 229 Ft./Minutes
Average Overland Flow Distance/Isochrone: 229 Ft./Min. x 60 Minutes = 13,750 Feet
NOTE: Above calculations are actuals for Sewee Creek Watershed near Decatur, Tennessee.
Fi gure 15. Time Area Histogram Calculation (Typical)
5
Area: 35.7 km 2
.400 Concentration Time: 75 Min. WATERSHEDAverage Velocity: 180 m/Min. BOUNDARY
Maximum Stream Length: 13,500 MetersDelta Elevation: 1,325 Meters
.300 -ISOCHRONE
3
o w .200
.100STREAM GAGE
1 2 3 4 5
0 15 30 45 60 75
Time (Minutes)
Figure 16. Typical Watershed (Noland Creek, N. C.) Divided by Isochrones
Table 6. Time Area Histogram Worksheet (Typical)
Isochrone No. Area (Sq. Miles) Fraction of Total
1 3.9 0.034
2 16.9 0.144
3 6.8 0.058
4 11.0 0.094
5 18.3 0.156
6 23.2 0.198
7 15.3 0.131
8 21.6 0.185
117.0 1.000
NOTE: Above values are actuals for Sewee Creek Watershed near Decatur, Tennessee
3.1.6 WATER SURFACE FRACTION (FWTR)
This is a fraction of the total watershed area covered by watersurfaces at normal low flow conditions. The value is estimated fromaerial photos and is virtually zero for watersheds containing neitherlakes nor swamps.
3.1.7 VEGETATIVE INTERCEPTION MAXIMUM RATE (VINTMR)
VINTMR is the maximum rate of rainfall interception by the water-shed vegetation expressed in inches per hour. The values range from0. 10 for grasslands to 0. 20 for heavy forest and is estimated from landcover interpretation from aerial photographs. A weighted average isused for watersheds having more than one type of vegetative cover.See reference 13.
3.1.8 OVERLAND FLOW MANNING'S "N" FOR PERVIOUS SURFACES(OFMN)
This is a roughness coefficient for overland flow derived frompublished tables dependent on estimated vegetative cover and soilusage. Weighted averages are used where different types of cover arein evidence. Values range from 0. 018 for smooth earth to 0. 100 forheavy forest. (See Table 7 for more details.)
3.1.9 OVERLAND FLOW MANNING'S "N" FOR IMPERVIOUS SURFACES(OFMNIS)
Derivation is the same as for OFMN but applies to impervious surfaces.Values are derived from published tables and vary from 0. 013 for smoothasphalt to 0.017 for unfinished concrete. (See Table 7.)
3.1.10 GROUND WATER EVAPOTRANSPIRATION FACTOR (GWETF)
GWETF is a factor used to estimate the current rate at whichswamp vegetation draws water from below the water table. This valueis usually zero unless aerial photographs reveal a significant area ofswamps within the watershed perimeter. In this case an appropriatevalue may be estimated using methods analogous to that used indetermining FIMP.
3-16
Table 7. Manning's Roughness Coefficient for OverlandFlow for Various Surface Types
Watershed Surface Manning's "N"
Smooth Asphalt 0.013
Concrete (Trowel Finish) 0.013
Rough Asphalt 0.016Concrete (Unfinished) 0.017Smooth Earth 0.018Firm Gravel 0.020
Cemented Rubble Masonry 0.025Pasture (Short Grass) 0.030Pasture (High Grass) 0.035Cultivated Area (Row Crops) 0.035Cultivated Area (Field Crops) 0.040
Scattered Brush, Heavy Weeds 0.050Light Brush and Trees (Winter) 0.050Light Brush and Trees (Summer) 0.060Dense Brush (Winter) 0.070Dense Brush (Summer) 0.100Heavy Timber 0.100
SOURCE: Reference 15, pp. 110-113.
3.1.11 SUBSURFACE WATER FACTOR (SUBWF)
SUBWF is the fraction of moisture entering groundwater storagewhich leaves the basin through subsurface flow (underground rivers,subterranean caverns) not measured by the stream gage. This value isusually zero unless such subterranean geophysical anomalies are knownto exist.
3.2 HISTORICAL DATA BASE
The historical data required by the simulation and calibrationmodels is similar in format to the physiographic data described inparagraph 3. 1. The basic difference is that while all physiographicdata is either measured or inferred from observations, historical datais converted from existing written records to a digital format for inputto the models. Where precipitation gages are not uniformly distributedabout a given watershed and are a combination of hourly and daily gages,Thiessen analysis is used to derive "weighting factors" to apply tothe historical data.
The historical data base generated by IBM for this hydrologyapplication study is constructed primarily from five different specifictypes of data:
1. Precipitation data
2. Stream stage charts
3. Stream gage rating tables
4. Daily discharge data
5. Evaporation data
The conversion of this data to the digital forms usable by the simulationand calibration models is represented in Figure 17. Table 8 is a listingof the historical data collected and presently on file in IBM, and Table 9is a summary of the utilization of that data base as applied to theinitial 15: watersheds being studied. The following paragraphs describein some detail the procedures used to extract the desired information fromthe historical data base.
3-18
PRECIPITATIONPRECIPITATI DATA
TABLESHOURLY
MERGE & SYNC IPTA1N
CHARTHARG
RATIN PA-MTDATA
PECIPI-
E TATION
DATA
FLOOD
HYDROGRAPH
PARAMETERS TORATING
TABLESTABLESTASKS 3 & 9ODAILY
DISCHARGEDAILY
TABLES
DAILY
DISCHARGE
DATA
EVAPORATIONTABLES
EVAPORATION
DATA
Figure 17. Task 2: Historical Data Base
Table 8. "Raw" Historical Data Acquired
(Tennessee Valley Region)
NWS DAILY CLIMATOLOGICAL DATA TABLES, 1948-70
TVA DAILY PRECIPITATION TABLES, 1948-70
NWS HOURLY PRECIPITATION DATA TAPES (5 REELS), 1960-70
SNWS DAILY PRECIPITATION DATA TAPES (9 REELS), 19637
TVNWS DAILY PRECIPITATION DATA TAPES (9 REELS), 1963-70
TVA DAILY PRECIPITATION DATA TAPE (1 REEL), 1968-71
USGS DAILY SURFACE WATER RECORDS, 1961-70
TVA DAILY DISCHARGE TABLES, 1940-70
USGS DAILY DISCHARGE DATA TAPE (1 REEL), 1939-70
SELECTED STREAM STAGE RECORDER CHARTS
Table 9. Processed Historical Data
MEAN BASIN/HOURLY PRECIPITATION* 15 MERGED RECORDS* 100 WATER YEARS* 294 WATER YEARS INPUT DATA
DAILY DISCHARGE* 15 RECORDS* 100 WATER YEARS
STORM EVENT HYDROGRAPHS
* PEAK DAY, HOUR, MAGNITUDE* 500 SIGNIFICANT STORMS
EVAPORATION* 15 RECORDS* 100 WATER YEARS
3.2.1 DATA AVAILABILITY SUMMARY
Precipitation data is gathered throughout the Tennessee Valley bymany precipitation stations which are owned and/or managed by either
TVA, the National Weather Service (NWS), or private corporations. Inall cases the data is recorded either hourly or daily and the informationstored on magnetic tape and/or written documents. Magnetic tapestorage is limited primarily to data collected from certain stations since1960; therefore, most precipitation data of value to the study is recordedin tables. The following paragraphs detail the step-by-step proceduresnecessary to obtain the hourly precipitation data required by the modelsfor a given watershed.
Once a subject watershed has been identified, several initialquestions must be answered:
1. How many precipitation stations exist within the boundariesof the watershed ?
2. How many precipitation stations are closely adjacent tothe watershed and how close are they?
3. How many of the stations are hourly and how many are daily ?
4. What are the years of operation for the stations ?
5. How much of the existing historical data exists as writtentables and how much on magnetic tape ?
The answers to the above questions exist in different reports, lists, maps,documents, etc., and require considerable manual effort to compile.Figure 18 is a sample of a matrix type data worksheet developed tocorrelate the answers to the questions. The following paragraphs explainthe notation used in Figure 18.
The watershed name, approximate latitude and longitude boundaries,centroid, and station names are self-explanatory. The station descriptionsare (1) S.G. for stream gage (required to determine the operational yearsto be selected for input to the simulation model), (2) PG-H and PG-D forprecipitation gage-hourly and daily, respectively, and (3) E. S. forevaporation station.
3-22
WATERSHED NAME: NOLAND CREEK NEAR BRYSON CITY, NORTH CAROLINA (13.8 SQ. MILES)
WATERSHED BOUNDARIES (APPROX.) LATITUDE 3528 TO 3534 ; LONGITUDE 8327 TO 8331WATERSHED CENTROID (APPROX.) LATITUDE 3531 ; LONGITUDE8329
S T A T I N N A M E Station TVA NWS LOCATION Read Data OPERATIONAL YEARS
Descr. No. No. Lat. Long. Time Type 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 6465 66 7 1 1
NOLAND CfEEK S.G (3-5135) 3529 8330 R T/D
NOLAND CREEK PG-H 183 3529 8330 R D *
NEWFOUND GAP PG-H 819 3535 8327 R D
GATLINBURG PG-H 209A 3420 3541 8332 R D 0
TOWNSEND PG-H 715A 3538 8345 R D
CADES COVE PG-D 177A 3534 8348 0900 D
GATLINBURG PG-D 209 3540 8330 0990 D
MT. LE CONTE PG-D 210 3535 8328 0900 D
ELA PG-D 525 3529 8325 0900 D
JEFFERSON CITY, TENN E.S. 3607 8330
Hourly Discharge Data Available.
Figure 18. Historical Data Availability
The station latitude and longitude are important in determining theproximity of the precipitation gages to the watershed. The location of the
various gages is required for the merging and synchronizing process
described in paragraph 3.2.2. The read time is either "R" for continuous
recording or is specified for daily stations (e.g., 2400 means the gage isread daily at midnight).
The data type is either "T" meaning magnetic tape and/or "D" for
written documentation. In those instances where both types are indicated,the upper line of the adjacent bar chart represents those operational yearswhere the data exist on magnetic tape. The lower line of the bar chartreflects the years that the station has been in operation and for whichprecipitation data (tape and/or table) is available. The "hourly dischargedata available" entry at the bottom of the charts represents those yearsfor which strip chart hydrographs are available.
Once this worksheet is completed, the water years to be used forthat particular basin can be selected, utilizing available stream gagedata and the best of the available precipitation gages. After selectionof operational water years and precipitation stations, conversion of theraw data to the digital data required by the models can begin.
3.2.2 HOURLY PRECIPITATION DATA
Hourly precipitation data in digital form is the primary input toKWM and OPSET. In a very small watershed having its own hourlyprecipitation gage (e. g., White Hollow) one can with reasonable safetyassume that the gage reading applies uniformly to the entire watershed.This assumption (which is implicit in both programs) departs from realitymore and more with increase in watershed size. It has been necessary,as part of the study, to implement a method whereby several precipitationrecords are used to synthesize a single hourly rainfall history for eachbasin.
The number of precipitation stations associated with any givenwatershed may vary from one station located 20 or 30 miles from thewatershed centroid to 5 or 6 stations located within or closely adjacentto the watershed boundaries.
3-24
Typically, a watershed will have one or two hourly stations, and -
one or more daily stations. In addition to the varying distances of these
stations from the centroid, the reading time for the daily stations might
be different. It is also quite likely that data will appear from the several
gages in both magnetic tape and tabular formats. The latter must be
manually extracted from the tables and converted to punched data card
format.
The precipitation gage outputs are assigned weighting factors, using
the Thiessen technique 2 , 15, in accordance with their physical locations
relative to the basin centroid. A software program, designated MERGE
AND SYNC in Figure 17 which was developed by IBM Huntsville, auto-
matically performs the interpolation and correlation of the precipitation
data. This program accepts all precipitation data, the reading time for
each daily station, and the weighting factor developed from the Thiessen
Analysis, and produces an hourly precipitation record for the applicable
water years associated with a given watershed. This hourly precipitation
data record is then used as one of the climatological inputs required by
the models.
3.2.3 SELECTED STORM DATA
For operation of the OPSET program it is necessary to select up to
five flood hydrographs for each of the years for which the model is to be
calibrated. This requires a manual search of precipitation and discharge
records to select storms useful to the calibration. The digitized input
data include the number of hydrographs chosen and three parameters
related to each hydrograph: day of occurrence of the flood peak, hour of
occurrence of the flood peak, and flow rate at the peak. These hydrographs
parameters are essential for the OPSET program to determine watershedmodel routing parameters, so that total flows will represent accurate
predictions, with respect to the time of occurrence of hydrograph peaks
as well as the total volume of flow for a given period of time. In practice
the selected storm hydrograph parameters are not available in daily
discharge records. It is necessary to obtain them from the strip charts
produced by the stream gage recorders. Rating tables are also digitizedand stored for conversion of gage height readings into flow rate.
The procedure employed to obtain this data requires manual analysis
of each strip chart and manual recording of the rise and fall of the stream
3-25
gage on an hourly basis. The time frame should extend from midnightof the day in which the storm occurred until some time at which thestream height returns to or approaches its initial stage. This hourlyheight recording is then formatted for entry into the computer where asubroutine will fetch the appropriate rating table into memory andconvert the data to cubic feet per second. This flood hydrograph datais then in a usable form when required by the simulation model.
3.2.4 DAILY DISCHARGE DATA
Daily discharge data is the volume in cubic feet of water per daythat flows past any given stream gage. This data exists on magnetictape and/or written tables for all stream gages in the Tennessee Valley.The data format which exists on magnetic tape must be altered to becompatible with the simulation model. Where the data exists in writtentables, it is necessary to manually extract that information, convert topunched card format, and develop a listing compatible with model requirements.
3.2.5 EVAPORATION TABLES
Evaporation data appear in Climatological Data publications of theNational Weather Service. Unfortunately, the number of pan evaporationstations is too limited to provide complete coverage. The nearest evapo-ration station may be as much as 100 miles from the watershed. Additionally,the station may be associated with a large lake or reservoir which hasevaporation rates different from those of an interior watershed in a pre-dominantly mountainous region. Preparation of the evaporation data issimilar to that for daily discharge data in that the rates and pan evapo-ration coefficients are read from published tables, punched onto cards,and a computer-compatible listing generated for the identified watershed.
3.3 MASTER WATERSHED DATA BANK
The digital products of Task 1 (Physiographic Data Base) and Task 2(Historical Data Base) are stored in a Master Watershed Data Bank, whichis in a digital tape format, along with control options and other logicalinputs needed for the operation of OPSET and KWM as indicated in Figure 19 .This tape-based data bank provides flexibility in operating the models andobviates storage of some 1,200,000 punched cards.
3-26
SIMULATIONFROM MODELTASK1 PARAMETERS
CONTROL OPTIONS
HOURLYPRECIPITATION rDATA
FROM EVAPORATION MASTER TODATA WATERSHED TASKS
DATABANK 4,5,6,7,9
DAILYDISCHARGEDATA
* Up to 50 WatershedsSELECTED e Up to 10 Years of DataSTORM for Each Watershed
Figure 19. Task 3: Master Watershed Data Bank
Considerable effort has been devoted to acquiring and formattingthe raw data and creating this master data bank. It is a by-product ofthe study which is potentially valuable to others engaged in hydrologicor related research using the same geographic area.
3.4 CALIBRATION (TASK 4)
3.4.1 USE OF OPSET
The OPSET model is used to calibrate a set of watersheds chosenfrom a region, generally using three water years of historical data foreach basin, determining a set of model parameters from the calibrationruns, and testing the accuracy of the calibration using the KWM. Cali-bration of several watersheds using OPSET generated sets of modelparameters that produced accurate simulations on a daily flow andmonthly flow basis which unfortunately did not accurately simulate thelow flows. The subroutine that estimates the Base Flow RecessionConstant (BFRC) has been modified, and results indicate an improvementin accuracy of simulation of low flows. Some manual "fine tuning" hasbeen used to obtain a set of model parameters that represents the water-shed very well.
Figure 21 shows the activities involved in Task 4 (Calibration). Datapertaining to the watershed to be calibrated are fed into the OPSET program,which is then run to estimate a set of model parameters. This step gets thetask "into the ballpark. " A simulation is then run using KWM and the IBManalysis/evaluation routines. Simulation accuracy is evaluated withrespect to total annual runoff, monthly flow, daily flow, statistical indices,and selected storm hydrograph characteristics. Based on these evaluations,parameters are adjusted, and a new simulation run, followed by anotherevaluation. This process goes through several iterations until simulatedflow matches observed flow with acceptable accuracy in all criteria ofinterest to the analyst. The choice of parameters to adjust, directionand magnitude of the adjustments depend upon the judgment of the analyst.
3-28
PRELIMINARY "OPTIMUM" MODEL TO-- SIMULATION EVALUATION
CALIBRATION PARAMETERS I TASKS 6, 9
OPSET KWM
DATA
FROM TASK 3 --
MANUAL SENSITIVITY
ADJUSTMENTS ANALYSIS
L (TASK 5)
Figure 20. Task 4: Calibration
Sensitivity analyses performed to date have produced invaluable guidanceto the manual-adjustment activity, reducing the subjectivity and eliminatingthe requirement that the analyst be skilled in hydrology.
3.4.2 PARAMETERS ESTIMATED BY CALIBRATION
The calibration process (a combination of automatic and manualadjustments as summarized above) was used to quantify the followingsimulation model parameters for each of ten watersheds. See Reference13 for more complete description.
1. BFRC, Base Flow Recession Constant, governs the rateat which groundwater flow recedes in the model.
2. IFRC, Interflow Recession Constant, governs interflowrecession.
3. BUZC, Basic Upper Zone Capacity, is an index forestimating the capacity of the soil surface (upper zone)to store water in interception and depression storage.
4. SUZC, Seasonal Upper Zone capacity adjustment constant,is used to adjust upper zone variations in vegetation andcultivation.
5. LZC, Lower Zone Capacity, is an estimate of the capacityof the basin soil to hold water. Decreasing LZC in themodel has the effect of increasing synthesized runoff.
6. BMIR, Basic Maximum Infiltration Rate, is the index usedto control the basic rate of moisture inflitration. This isa parameter to which simulation accuracy is very sensitive,particularly as it affects storm peaks.
7. SIAC, the Seasonal Infiltration Adjustment Constant, isan evaporation-infiltration factor relating infiltration ratesto evaporation rates to account for more rapid infiltrationduring warmer periods.
8. ETLF, Evapotranspiration Loss Factor, is an index usedto estimate the maximum rate of evapotranspiration whichcould accur within the watershed under current conditionsof soil moisture content.
9. BIVF, Basic Interflow Volume Factor, controls time distri-bution and quantities of moisture entering interflow. Increasing
3-30
BIVF tends to reduce storm runoff peaks and extendhydrograph recession limbs.
10. NCTRI, Number of Current Time Routing Increments, isthe number of subareas into which the basin should bedivided, given 15 minute or one hour separation betweenisochrones.
11. CSRX, Channel Storage Routing Index, is used to accountfor channel storage when channel flows are less than halfof channel capacity (CHCAP). Channel storage effectsare simulated by having the hydrograph time routed to themouth of the watershed through an imaginary reservoir.
12. FSRX, Flood plain Storage Routing Index, is used to accountfor channel storage plus flood plain storage when stream-flows are greater than twice the channel capacity. Betweenone-half and twice channel capacity, the program inter-polate values between FSRX and CSRX.
13. CHCAP, Channel Capacity, is that value of streamflow,measured at the gage, at which a transistor is made fromchannel routing to flood plain routing. In mountainouswatersheds, this is not an oritical parameter, and OPSETseldom adjusts it.
After OPSET has adjusted the parameters listed above to achieve a"best match" based on mean daily streamflow, it is usually found thatsynthesized flood peaks fail to match observed peaks, in magnitude and/ortime. Since the study attempted to address as wide a variety of applicationsas practicable, some manual "fine tuning" was undertaken to achievean acceptable match between synthesized and observed flood peaks whilemaintaining an acceptable correlation between synthesized and observedmean daily and monthly flows. This manual adjustment process requiressome knowledge of the hydrologic processes occurring in the watershedand some subjective judgment. No firm rules or recipes have beendeveloped, but the following are useful guidelines.
1. Overall results are affected by soil moisture capacitiesand infiltration rates (LZC, BUZC, BMIR) and their relatedseasonal adjustment constants (SUZC, SIAC).
2. Initial storages (LZS and UZS) can be varied to improveaccuracy in the first two months of the multiyear simula-tion.
3-31
3. Summer storm peaks are affected more than winterstorm peaks by changes in SIAC and SUZC, and thelatter has more influence on mean daily flow in driermonths than the former.
4. Consistent phasing errors (differences in times of occurrencebetween synthesized and observed flood peaks) can usuallybe reduced by adjustment if the number and sizes ofsubareas in the time-area histogram. Phasing errorswhich appear random are attributable to errors in inputprecipitation and/or evaporation data.
5. Since there are parameter interactions, all performanceindices should be re-checked after any parameter adjust-ment and others re-adjusted as needed until a "best simula-tion" is realized.
3.4.3 CALIBRATION RESULTS
The calibration process was applied to ten Tennessee Valleywatersheds to derive the sets of parameters shown in Table 10. Parameterdefinitions are listed in Table 11. Each set is the one which producessimulations that are acceptable with respect to all indices of performance,as sho*n in Table 12. The mean monthly and daily streamflows showgood correlation coefficients.
From each multi-year simulation, the largest winter storm wasincluded in the table, even though others showing more felicitous resultscould have been chosen. For some of the large winter storms (e.g.,Sewee Creek, 3-13-63), simulated flood peaks vary from the observedpeaks by factors in magnitude and several hours in time. Experiencein the study shows that (1) input data are often erroneous and (2) goodsimulation fidelity with respect to storm/flood events should not beexpected in a free-running or "open loop" model. In an operationalapplication, moisture storage conditions and simulated flow rate wouldbe periodically adjusted in accordance with actual observed conditions andin a short-term forecast (24 to 72 hours), a prediction accuracy withinfive per cent could confidently be expected.
Several plots were generated early in the study after the first cali-brations were completed. They were useful at first, but more time-consuming than the information they yielded would justify. It was found thatsimulated and observed storm events could be compared effectively, forevaluation purposes, on the basis of peak flow magnitude and time. Some
3-32
Table 10. Simulation Model Parameters After Calibration
WATERSHED PARAMETERS QUANTIFIED BY CALIBRATION SOILHWATERSHED CHANNEL ROUTIN CHARAC-
PARAMETERS SOIL MOISTURE AND GROUNDWATER OVERLAND FLOW TERISTICS
NO. NAME AREA FIMP FWTR BMIR LZC ETLF SUZC SIAC BUZC BIVF CSRX FSRX BFRC BFNLR OFSL OFSS OFMN IFRC OFMNIS AWC PERM-A
1 WHITE HOLLOW 2.68 .001 .001 9.0 6.0 .20 .50 .40 .90 .50 .99 .97 .98 .85 1000 .30 .10 .10 .015 3.71 1.70
2 LITTLE CHESTUEE 8.26 .013 0.0 10.0 6.0 .20 .20 .40 .50 0.0 .93 .97 .99 .85 630 .237 .061 .10 .002 8.75 1.20
3 SOUTH FORK MILLS 9.99 .001 0.0 12.9 7.6 .07 .20 .20 .20 .58 .96 .97 .99 .85 1140 .315 .05 .10 .015 7.90 4.54
5 WEST FORK PIGEON 27.6 .001 .008 4.0 7.0 .10 .15 .15 .15 0.0 .95 .95 .99 .85 1380 .413 .05 .10 .015 8.82 4.0
C 7 BIG ROCK CREEK 24.9 .033 .005 5.0 4.0 .07 .15 .15 .40 0.0 .94 .95 .70 .50 980 .155 .04 .10 .015 5.30 .77
9 TRACE CREEK 30.4 .039 .005 6.5 4.6 .11 .70 .90 .39 0.0 .97 .97 .96 .85 850 .129 .05 .10 .015 1.80 1.43
12 EMORY CREEK 83.2 .005 .001 7.0 3.9 .10 .20 .20 .30 .78 .94 .94 .84 .85 1270 .298 .07 .10 .010 5.55 2.67
13 SEWEE CREEK 117 .001 .001 4.0 3.5 .10 .25 .15 .25 .50 .98 .99 .99 .85 1060 .165 .05 .10 .015 4.92 .50
14 TOWN CREEK 141 .002 .001 7.0 4.0 .25 .25 .30 .25 .50 .94 .90 .93 .85 1550 .062 .05 .10 .014 6.19 4.3
15 POPLAR CREEK 82.5 .010 .001 4.0 3.0 .15 .20 .15 .15 1.0 .97 .98 .98 .85 1320 .329 .05 .10 .015 2.02 1.08
NOTE: Watershed parameters and soil characteristics are included in the table to make a complete base for correlation.
Table 11. Model Parameter Definitions
SYMBOL DEFINITION/SIGNI FICANCE
BIVF Basic Interflow Volume Factor; indexes distribution and quantities of moisture entering interflow.
BFNLR Base Flow Nonlinear Recession Adjustment Factor
BFRC Base Flow Recession Constant
BMIR Basic Maximum Infiltration Rate; relates to "A" horizon permeability.
BUZC Basic Upper Zone Storage Capacity
CHCAP Channel Capacity Indexed to Basin Outlet; flow rate indicated at transition from channelrouting to flood plain routing.
CSRX Channel Storage Routing Index; used to account for channel storage when channel flows are lessthan half of CHCAP.
ETLF Evapotranspiration Loss Factor; indexes moisture depletion rate through evapotranspiration.
FSRX Flood Plain Storage Routing Index; used to account for channel plus flood plain storage whenstreamflows are greater than twice CHCAP. The model synthesizes a routing index when
W streamflow is between 2*CHCAP and 2*CHCAP.
IFRC Interflow Recession Constant; indexes interflow recession rate, effective only in conjunctionwith a non-zero BIVF.
LZC Lower Zone Capacity; indexes lower zone soil moisture storage capacity; relates to plant-available water capacity.
NBTRI Number of Base Time Routing Increments; used by the model in routing by Clark's method.
SIAC Seasonal Infiltration Adjustment Constant; relates to changes in vegetative cover with changesin season.
SUZC Seasonal Upper Zone Storage Capacity Factor; BUZC and SUZC together regulate the capacityof the upper zone to hold moisture.
VINTMR Vegetative Interception Maximum Rate.
Table 12. Simulation Accuracy After Calibration
STREAMFLOW LARGEST WINTER STORMWATERSHED YEARS
OF CORRELATION DAILY, CFS PEAK
AREA RECORDS COEFFICIENT MEAN MAX DATE FLOW, CFS TIME, HRNO. SHORT NAMEMI2 KM 2 YRS FROM THRU M'LY D'LY OBS SIM OBS SIM OBS SIM OBS SIM
1 WHITE HOLLOW 2.7 6.9 7 57 63 .95 .91 3.79 3.69 147 78 3-12-63 270 98 6 4
2 LITTLE CHESTUEE 8.2 21.3 7 50 56 .92 .87 13.0 13.9 388 447 4-15-56 1,030 895 18 18
3 SOUTH FORK MILLS 9.9 25.7 3 66 68 .84 .80 32.2 32.5 713 706 2-13-66 1,618 1,273 6 8
5 WEST FORK PIGEON 27.6 37.3 9 56 64 .93 .82 78.1 79.1 1600 2641 3-5-64 4,714 3,368 25 27
7 BIG ROCK CREEK 24.9 64.5 7 55 61 .95 .93 40.3 40.9 3800 3895 3-7-61 6,518 5,923 24 26
9 TRACE CREEK 30.4 78.7 6 64 69 .93 .88 39.8 37.5 2310 1547 12-4-64 3,524 2,579 8 8
12 EMORY RIVER 83.2 215 4 53 56 .95 .83 142 145 5460 3150 3-22-55 13,000 8,925 3 1
13 SEWEE CREEK 117 303 9 60 68 .93 .84 192 142 6450 5658 3-13-63 20,400 8,561 35 46
14 TOWN CREEK 141 365 6 61 66 .95 .90 282 216 9260 8384 3-25-64 10,699 9,233 48 49
15 POPLAR CREEK 82.5 214 7 62 68 .96 .90 157 142 6200 4068 3-12-63 6,389 5,252 38 31
of the plots generated early in the study are shown in Figures 21 (whose
time scale is unavoidably too small for good interpretation) through 25.
Except for some graphical enhancement for the sake of good reproduction,all the plots except that of Figure 23 were generated automatically.
3-36
[II I I:II I Ill 1 I I I I I--L- ....
60 .:.1699f l = l I : I i J [ i .I. . ..l 1
. .
50 1416
0I- : 6
IL I-
... i
420 1 23
10 f
Figure 21. One Year Simulated and Observed Flow, White Hollow, 1961
DAYS~~~~~I FRMOTBRTROG1ETME
Figue 2. Oe Yar imultedandObsrve Flo, WiteHolow,196
70 1882
60 .:. , - - 1699
S 50:w w
w LU
CAD 850
C i I-
020 566
S= SIMULATED ;
O= OBSERVED
LI-
I-,
20 S S IM U LA T E D 566
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 0
MAR
Figure 22. One Month Simulated and Observed Flow, White Hollow, March 1961
(2.5) -.1
(5.0) -.2
(7.6) .3
(10.2)-.4
70- (1882) (12.7)-.5 6
(15.2) .6 Z
(17.8) -.760- (1699) (21.3) .
50- (1416) -------- OBSERVED
S\SIMULATED
U.O 40- (1133)
, 0
10-
20- (566)
10- (283)
10 20 30 40 50 60 70 80 90 100 110 120 130 140
TIME, HOURS
Figure 23. Selected Storm, November 16-23, 1957, White Hollow
-*LITTLE CHESTUEE, BELOW WILSON, TENN. (8.24 SQ.MI.) WY51 STUDY LC08 E
400 - 11.3
360 -10.2
3 320 -- 9.1o OBSERVED STREAMFLOW. --- SIMULATED STREAMFLOW
C280 7.9
200 5.7
I-
> 160 4.5
w
0
> 120 3.4
-J
z
80 2.3
0.00.0 4 8 12 16 20 24 28 32
MARCH (DAYS)
OBSERVED MONTHLY STREAMFLOW = 1301 CFS SIMULATED MONTHLY STREAMFLOW = 1156 CFS
Figure 24. One Month Simulated and Observed Flows, Little Chestuee,March 1951
3-40
0)*LITTLE CHESTUEE, TENN. 8.24 SM. SN5103 03/28/51-03/29/51 STUDY LC08
1000 31.2
900 _28.3
S80I0 / _ 25.5
OBSERVED STREAMFLOW- - - SIMULATED STREAMFLOW I
- ......... OBSERVED PRECIPITATION Iu. _22.72 700 22.7
I-
co60 19.8
' 600 '
-J
C 500 17.0
> I
> \400 14.2S,, \
0 \
S300\ 11.3
S/ \ I.
200 5.7
100 / 2.8
0.00.0 4 8 12 16 20 24 28 32 36 40 44
OBSERVED RUNOFF = 2.25 INCHES TIME (HOUR) SIMULATED RUNOFF = 2.81 INCHES
Figure 25. Simulated and Observed Runoff, Major Winter Storm,Little Chestuee, March 28-29, 1951
3-41
90.63200
OBSERVED
sweenww SIMULATED
79.22800
.67.9o 2400
0U..I
SObserved Total Streamflow = 22,289 CFS
w Simulated Total Streamflow = 18,799 CFS 56.6
a 2000
45.32 1600
z
> 33.9u - 1200
ca
-
800 22.7
400 '111.3
0.02 4 6 8 10 12 14 16 18 20 22 24 26 28 30
FEBRUARY (DAYS)
Figure 26. Simulated and Observed Flows, Emory River, February 1956
3-42
3.5 SENSITIVITY ANALYSIS (TASK 5)
-The success of-a-general hydrologic model is measured by itsability to simulate streamflow sequences that match observed records.The model utilizes time sequences of climatological data, informationon the physical watershed, and a set of values for model parameters.These model parameters theoretically relate to watershed physicalcharacteristics; operationally, those values are estimated by a sequenceof trials and adjustments ending in an acceptable match. The quality ofa given trial is determined by the closeness with which the observed andsimulated flows agree during every simulation period. The manifestimpracticality of making all these comparisons in evaluating a trialsimulation requires selection of a small number of quality indices.One such index is the total annual flow; however, a large number ofcombinations of parameter values will give the same total annual flow.Therefore we need to differentiate in selecting among these combinationsby adding other indices such as mean, standard deviations, root sumsquares, daily correlation coefficients, low and high flows, and hydro-graph characteristics.
If one is to adjust a trial set of model parameter values in orderto improve the matching of the observed and simulated streamflows, heneeds information on the effect a given change in a given parameter valuewill have on the simulated streamflows. The universe of such informationis a multidimensional response surface of values for each index at pointsrepresenting each set of parameter values. The response surface ofinterest is bounded by the reasonable range for each individual parameter.Comprehensive mapping is impractical.
The information on the direction and the rate of change of the indexvalues per unit change of each parameter at the point representing the setis important and very useful. Theoretically an infinite number of pointscould be analyzed, but a good one to start with is one accepted after trialand error modeling to achieve a "best" match.
The advantages of the sensitivity analysis lie in its ability to: (1)guide manual adjustments or programmed adjustments in a computerizedoptimization routine, indicating how much to change a given mode param-eter to effect a given correction in a statistical index and/or hydrographcharacteristic, and also the effect it will have on other indices; (2) showwhich parameters need to be estimated carefully and which ones requireonly rough approximation; (3) indicate the critical parameters on whicheffort should be concentrated to find a correlation with watershed physicalcharacteristics that can be observed by remote sensing; and (4) provideguidance on what difference in runoff characteristics to expect from agiven difference in physical characteristics of two watersheds.
3-43
Sensitivity analysis is time consuming and costly because of thelarge number of computer runs required. Also, a completed sensitivityanalysis may be invalidated by a change in the model, but recently themodels have been stabilized sufficiently to make the analysis worthwhile.Figure 27 shows the interfaces of Task 5 (Sensitivity Analysis) with otherstudy tasks.
Two sensitivity analyses were performed, for the purpose ofguiding manual calibrations, on two small basins: White Hollow andLittle Chestuee. The results of the latter are tabulated in Table 13.The column headings represent various mean daily streamflow statis-tics and hydrograph characteristics that were selected to assess the close-ness of simulated flow to recorded flow. No single statistical qualityindex can be used in reckoning the quality of a match. Different usersneed to match different statistical indices depending on their objective,interest, or application (e.g., flood control, low flows, water supply).None of the parameters except NBTRI affect flood peak timing. Floodpeak times are therefore not included in Table 13. Parameter valuesproducing the "best" simulation appear in parentheses in the PARAMcolumn.
The data shown in Table 13 need further interpretation to be in-formative. A "Unit Sensitivity", defined by the following division
US = (percent change in simulation result)/(percent changein parameter)
was calculated for each parameter with respect to the simulation resultof interest, as a measure of parameter effectiveness. The results forthe Little Chestuee basin are summarized in Table 14.
Examination of the results suggests a general scheme of parameterchange that should improve accuracy of low flow simulation, and the peakof the biggest storm. Increasing the BFRC value from 0. 98 to 0. 99 wouldimprove the summer low flow simulation. A reduction in ETLF from 0. 165to 0. 05 would increase the annual flow volume, and counteract the reductiondue to BFRC increase. Reducing the CSRX value from 0. 97 to 0. 96 shouldincrease the peaks of all floods and therefore better the peak of the biggeststorm. Results of experiments with these combinations were used to advan-tage in the calibration task.
3-44
"OPTIMUM" PARAMETERS GUIDANCE TO BEST CALIBRATIONFROM TASK 4 TO TASK'4
Sensitivity Analysis
PARAMETER REVIEW AND----SIMULATION -- :i TO TASKS 8, 9
PERTURBATIONS EVALUATION SIU- LATOiNi
KWMSENSITIVITY OFSIMULATIONACCURACY TOPARAMETER
VARIATIONS
Figure 27. Task 5: Sensitivity Analysis
Table 13. Sensitivity Analysis of Little Chestuee (Part 1 of 4)
SENSITIVITY ANALYSIS FOR. LITTLE CHESTUEE WATERSHED WATER YEAR 1951 DRAINAGE AREA 8.24 Snar Mile
DAILY STREAMFLOW STATISTICAL SUMMARY STORM ANALYSIS SUMMARY
STORM OF STORM OF STORM OF STORM OF
FLOW ERROR AVERAGE LOW FLOW 1/14 /51 2/ 1 / 51 3 / 29 /51 9/22 /51
STATISTICS STATISTICS ERRORS STATISTICS # #2 #3 #4
ROOT DAILY MEAN MEAN LOWEST PEAK R/O PEAK R/O PEAK R/O PEAK R/OA STA SUM LOW HIGH OF OF OF
M EVN OUR CORR. FLOWS FLOWS I)NIgMEAN DEV. SoAE COE F F S 33.1) Aug. IJun I Feb. (CFS) IHOUR) I N ) (CFS) OURI ( IN ) CS o ) (CF) uOUP i N
OBSERVEDICONFIGURATION 14.01 27-9 6.25 7.24 11.0 191 21 0.49 798 9 1.61 954 2 2.25 230 15 .20
BEST SIMULATEDCONFIGURATION 14.78 33.67 194 .963 +0.7 +1.6 4.50 10.97 9.7 222 19 0.61 950 8 1.86 954 4 2.83 231 16 .28
I D PARAM VALUE WATERSHED PARAMETERS
51 AREA 4.12 7.38 16.83 271 .963 -2.5 -21.7 2.24 5.46 4.9 111 19 .30 475 8 .93 477 4 1.41 115 16 .14
52 (8.24) 12.36 22.17 50.41 498 .963 +3.9 +24.9 6.74 16.46 14.6 333 19 .91 1424 8 2.80 1432 4 4.24 346 16 .43
53 FIMP 0.00 14.59 33.33 188 .963 +0.6 +0.2 4.36 10.71 9.9 '219 19 .60 944 8 1.85 950 4 2.80 216 16 .26
C4 57 (0.013) 0.052 15.36 34.74 216 .959 +0.9 +5.8 4.90 11.76 9.4 232 19 .64 966 8 1.91 969 4 2.90 273 16 .35
58 0.10 16.07 36.17 249 .952 +1.1 +11.0 5.39 12.75 9.1 244 19 .68 987 8 1.98 987 4 2.99 326 16 .43
59 0.20 17.56 39.48 330 .930 +1.7 +21.8 6.42 14.81 8.1 269 19 .75 1030 8 2.10 1023 4 3.19 437 16 .55
60 0.50 22.04 51.20 612 .852 +3.2 +54.0 9.54 21.15 5.5 343 19 .99 1158 8 2.48 1134 4 3.77 775 16 1.09
63 FWTR 0.01 14.67 33.95 198 .962 +0.4 +2.5 4.09 10.71 9.6 224 19 .61 954 8 1.88 958 4 2.84 240 16 .30
65 (0.0) 0.10 14.30 36.44 248 .952 -1.1 +10.1 3.34 8.95 8.2 244 19 .67 989 8 1.97 989 4 2.99 327 16 .42
67 0.50 18.20 49.27 557 .862 +0.4 +44.3 7.24 15.27 2.4 332 19 .93 1144 8 2.41 1128 4 3.567 717 16 .98
SOIL MOISTURE PARAMETERS
2 8MIR 3.15 15.31 41.36 339 .942 -0.5 +18.0 3.44 10.59 6.8 307 19 .76 1109 8 2.25 1099 4 3.38 356 16 .43
3 (6.3) 9.45 14.53 28.36 132 .970 +1.4 -6.6 5.08 11.54 11.6 165 19 .51 803 8 1.57 825 4 2.38 161 16 .21
6 LZC 4.00 16.67 41.42 328 .954 +0.3 +16.8 4.29 11.57 7.4 254 19 .83 1045 8 2.30 1044 4 3.24 199 16 .25
7 18.0) 12.00 12.59 28.21 157 .958 0.0 -7.8 4.50 10.64 8.9 159 19 .45 814 8 1.51 856 4 2.43 274 16 .34
10 ETLF 0.125 15.42 34.26 211 .958 +1.1 +6.3 5.18 11.68 9.7 225 19 .62 957 8 1.89 960 4 2.85 420 16 .51.
11 (0.25) 0.375 14.5 33.51 192 .963 +0.5 -0.4 4.24 10.70 9.8 221 19 .60 947 8 1.86 953 4 2.82 141 16 .18
14 SIAC 0.15 14.77 31.81 171 .964 +0.8 +0.2 4.59 11.58 10.7 200 19 .57 891 8 1.74 909 4 2.66 256 16 .31
15 (0.30) 0.45 14.80 35.41 221 .961 +0.6 +3.3 4.40 10.37 9.1 244 19 .65 1001 8 1.98 995 4 2.97 205 16 .26
18 SUZC 0.15 15.32 33.94 211 .955 +1.0 +3.8 5.40 13.24 9.8 222 19 .61 949 8 1.86 954 4 2.82 328 16 .40
19 (0.301) 0.45 14.32 33.77 191 .965 +0.3 -1.2 4.03 9.76 9.6 225 19 .61 955 8 1.88 959 4 2.84 138 16 .18
22 BUZC 0.21 14.81 33.68 195 .963 +0.7 +1.9 4.53 11.02 9.7 222 19 .61 949 8 1.86 954 4 2.82 244 16 .30
23 (0.42) 0.63 14.75 33.67 194 .963 +0.7 +1.4 4.46 10.91 9.7 222 19 .61 949 8 1.86 955 4 2.83 216 16 .27
40 BIVF 0.20 14.77 33.78 205 .957 +0.7 +1.7 4.51 11.05 9.8 229 19 .60 965 8 1.87 972 4 2.87 231 16 .29
41 (0.40) 0.60 14.82 33.21 166 .975 +0.5 +1.6 4.37 10.95 9.5 173 19 .61 855 8 1.84 900 4 2.70 229 16 .28
68 VINTMR 0.00 14.88 33.18 184 .965 +1.0 -0.2 4.77 11.32 10.2 208 19 .56 939 8 1.82 955 4 2.80 230 16 .29
69 10.149) 0.075 14.79 33.53 191 .963 +0.8 +1.0 4.54 11.03 9.9 219 19 .59 947 8 1.85 954 4 2.82 230 16 .29
71 0.224 14.76 33.82 197 .962 +0.7 +1.9 4.46 10.95 9.7 225 19 .62 954 8 1.88 955 4 2.4 231 16 .28
95524262
Table 13. Sensitivity Analysis of Little Chestuee (Part 2 of 4)
SENSITIVITY ANALYSIS FOR, LITTLE CHESTUEE WATERSHED WATER YEAR 1951 DRAINAGE AREA 8.24 Square Miles
DAILY STREAMFLOW STATISTICAL SUMMARY STORM ANALYSIS SUMMARY
STORM OF STORM OF STORM OF STORM OF
FLOW ERROR AVERAGE LOW FLOW 1 / 14 / 51 2 / 1 / 51 3/
29 / 51 9/ 22/ 51STATISTICS STATISTICS ERRORS STATISTICS #1 #2 #3 #4
ROOT DAILY MEAN MEAN LOWEST PEAK R/O PEAK R/O PEAK R/O PEAK R/OMEAN ST SUM CORR. LOW HIGH OF OF OF
DEV. UARE COEFF 14.5 33 Aug. June I I Feb. C I FSI o n INI ICFS) H* UR)I I N I (CFSI IHOu l I I N (CFSI ''HO.' I N
OBSERVEDCONFIGURATION 14.01 27.9 5.25. 7.24 11.0 191 21 0.49 798 9 1.61 954 2 2.25 230 15 .20BEST SIMULATEDCONFIGURATION 14.78 33.67 194 .963 +0.7 +1.6 4.50 10.97 9.7 222 19 0.61 950 8 186 954 4 2.83 231 16 .28
I D PARAM VALUE SOIL MOISTURE PARAMETERS (CONTINUED)
72 0.298 14.75 33.98 200 .961 +0.6 +2.0 4.43 10.90 9.6 228 19 .63 958 8 1.89 955 4 2.85 231 16 .28
74 1.00 14.64 34.56 214 .958 +0.4 +2.1 4.20 10.66 9.3 228 19 .63 964 8 1.91 962 4 2.92 234 16 .29
122 SUBWF 0.05 14.38 33.60 194 .962 +0.4 +1.0 4.28 10.56 9.3 222 19 .60 949 8 1.86 954' 4 2.82 230 16 .28
123 (0.0) 0.10 13.97 33.52 193 .962 +0.1 +0.4 4.07 10.14 8.8 221 19 .60 948 8 1.85 953 4 2.81 230 16 .28
124 0.20 13.17 33.38 194 .961 -0.4 -0.9 3.64 9.32 8.0 221 19 .59 947 8 1.84 951 4 2.80 230 16 .28
125 0.30 12.37 33.25 197 .960 -1.0 -2.1 3.21 8.50 6.8 220 19 .58 946 8 1.83 949 4 2.79 230 16 .28
126 GWETF 0.05 13.53 33.77 194 .964 -0.8 +0.7 2.31 8.50 9.3 222 19 .60 949 8 1.86 954 4 2.82 229 16 .28
127 (0.0) 0.10 12.72 33.80 195 .963 -1.6 -0.0 1.34 6.94 8.8 221 19 .60 949 8 1.86 953 4 '2.81 229 16 .27
128 0.20 11.68 33.77 199 .964 -2.6 -1.1 0.63 5.26 8.0 221 19 .59 948 8 1.85 952 4 2.81 228 16 .27
129 0.30 11.01 33.71 202 .963 -3.1 -2.0 0.41 4.46 7.4 220 19 .58 948 8 1.84 951 4 2.80 228 16 .27OVERLAND PARAMETERS
46 IFRC 0.05 14.78 33.77 197 .962 +0.7 +1.8 4.49 10.97 9.7 223 19 .61 952 8 1.87 959 4 2.84 231 16 .28
47 (0.10) 0.15 14.78 33.58 192 .963 +0.7 +1.5 4.49 10.97 9.7 222 19 .61 948 8 1.86 951 4 2.81 231 16 .28
77 OFSS 0.118 14.75 33.49 190 .964 +0.8 +0.9 4.56 10.89 9.9 219 19 .60 950 8 1.85 958 4 2.82 230 16 .28
78 (0.237) 0.355 14.79 33.77 197 .962 +0.7 +2.1 4.46 11.02 9.7 224 19 .61 950 8 1.87 953 4 2.83 231 16 .2979 0.474 14.81 33.83 198 .962 +0.6 +2.3 4.44 11.03 9.6 225 19 .61 950 8 1.87 952 4 2.83 231 16 .2981 0.948 14.83 33.98 202 .961 +0.6 +3.0 4.39 11.06 9.5 227 19 .62 951 8 1.88 951 4 2.84 231 16 .29
82 OFSL 1.0 14.98 34.45 212 .958 40.4 +5.5 4.26 11.10 9.1 232 19 .64 952 8 1.90 951 4 2.86 231 16 .2983 1630.0) 315.0 14.83 33.98 202 .961 +0.6 +3.0 4.39 11.06 9.5 227 19 .62 951 8 1.88 951 4 2.84 231 16 .2984 945.0 14.74 33.46 189 .964 +0.8 +0.7 4.57 10.88 9.9 218 19 .60 950 8 1.85 959 4 2.82 229 16 .2885 1260.0 14.71 33.28 185 .965 40.8 -0.2 4.63 10.85 10.0 214 19 .59 951 8 1,84 962 4 2.81 216 16 .27
87 3150.0 14.60 32.56 169 .969 +1.0 -3.7 4.86 10.77 10.5 187 19 .54 948 8 1.80 966 4 2.78 164 16 .2188 OFMN 0.030 14.83 33.99 202 .961 +0.6 +3.0 4.39 11.06 9.5 227 19 .62 951 8 1.88 951 4 2.84 231 16 .29
89 (0.061) 0.091 14.75 33.46 189 .964 +0.8 +07 4.57 10.88 9.9 218 19 .60 950 8 1.85 959 4 2.82 229 16 .2892 0.305 14.60 32.56 169 .969 +1.0 -3.7 4.86 10.77 10.5 187 19 .54 948 8 1.80 967 4 2.78 164 16 .2193 OFMNIS 0.0005 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 2.83 231 16 .2897 10.002) 0.01 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
955-2426-2
Table 13. Sensitivity Analysis of Little Chestuee (Part 3 of 4)
SENSITIVITY ANALYSIS FOR, LITTLE CHESTUEE WATERSHED WATER YEAR 1951 DRAINAGE AREA 8.24 Square Miles
DAILY STREAMFLOW STATISTICAL SUMMARY STORM ANALYSIS SUMMARY
STORM OF STORM OF STORM OF STORM OF
FLOW ERROR AVERAGE LOW FLOW 1/14 51 21 1 / 51 3 2951 9/ 22 151STATISTICS STATISTICS ERRORS STATISTICS #1 #2 #3 #4
ROOT DAILY MEAN MEAN LOWEST PEAK R/O R/O PEAK R/O PEAK R/O PEAK R/OMEAN STO SUM CORR LOW HIGH OF OF OF
S U E C FLOWS FLOWSDEV. VUARE COEFF FLOWS FLOW I jn I I Feb. I (CFS) IHOURI I N (CFSI (ONUR I IN I (CFSI oUR I I N ) (CFS) OlO I N
OBSERVED
CONFIGURATION 14.01 27.9 5.25 7.24 11.0 191 21 0.49 798 9 1.61 954 2 2.25 230 15 .20BEST SIMULATEDCONFIGURATION 14,78 33.67 1 194 .963 +0.7 41.6 4.50 10.97 9.7 222 19 0.61 950 8 1.86 "954 4 2.83 231 16 .28
I D PARAM VALUE CHANNEL ROUTING AND GROUNDWATER PARAMETERS
28 CSRX 0.925 14.81 34.28 218 .953 +0.7 +2.7 4.50 10.98 9.7 294 19 .61 979 8 1.88 962 4 2.90 312 16 29
29 (0.95) 0.975 14.76 31.87 133 .982 +0.7 -1.4 4.48 10.91 9.8 132 19 .59 637 8 1.77 742 4 2.49 128 16 .232 FSRX 0.925 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 2.83 2316
33 (0.95) 0.975 14.82 34.06 182 .972 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 729 8 1.86 781 4 2.78 230 16 .2936 BFRC 0.97 14.97 34.06 195 .965 +0.4 +3.1 3.17 9.75 10.4 223 19 .62 952 8 1.88 958 4 2.84 230 16 .28
37 (0.98) 0.99 13.87 33.11 198 .956 +0.8 -1.2 6.63 11.92 7.7 220 19 .58 946 8 1.84 949 4 2.79 232 16 .30
99 CHCAP 50.0 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 .2.83 231 16 .28
103 (200.0) 300.0 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
104 400.0 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
130 EXOPV 0.01 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
131 (0.30) 0.15 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
132 0.45 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
134 1.00 14.78 33.67 194 .963 +0.7 +1.6 4.49 10.97 9.7 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
135 8FNLR 0.1 14.83 33.97 194 .965 +0.4 +2.7 3.90 10.11 10.0 223 19 .61 951 8 1.87 957 4 2.84 230 16 .28
136 10.85) 0 14.80 33.83 194 .964 +0.5 +2.2 4.18 10.52 9.9 222 19 .61 950 8 1.87 956 4 2.83 230 16 .28
137 0.5 14.79 33.75 194 .963 +0.6 +1.9 4.33 10.73 98 222 19 .61 950 8 1.87 955 4 2.83 230 16 .28
138 0.5 0.7 14.78 33.70 194 .963 +0.7 +1.7 4.43 10.88 9.8 222 19 .61 950 8 1.87 955 4 2.83 231 16 .28
139 0.9 14.78 33.66 194 .963 +0.7 +1.6 4.51 10.99 9.7 222 19 .61 950 8 1.86 954 4 2.82 231 16 .28
STARTING MOISTURE STORAGE VALUES AS OF OCTOBER 1
141 GWS 0.01 14.57 33.71 195 .963 +0.4 +1.5 4.49 6.51 4.64 222 19 .60 949 8 1.86 954 4 2.83 231 16 .28
142 (10.355) 0.18 14.67 33.69 194 .963 +0.5 +1.6 5.05 6.82 4.49 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
143 0.54 14.89 33.65 194 .963 +0.9 +1.7 6.26 7.46 4.64 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
144 0.71 14.99 33.64 194 .963 +1.0 +1.7 6.82 7.77 4.64 222 19 .61 950 8 1.87 954 4 2.83 231 16 .28
145 UZS 0.1 14.81 33.76 196 .963 +0.7 +1.8 5.70 7.17 4.65 223 19 .61 952 8 1.87 956 4 2.83 231 16 .29
146 (0.0) 0.5 14.96 34.14 201 .962 +0.7 +2.7 5.97 7.32 4.67 227 19 .62 959 8 1.90 962 4 2.86 233 16 .29
147 1.0 15.22 34.33 205 .962 +1.0 +3.2 7.09 7.92 4.53 229 19 .63 964 8 1.91 965 4 2.87 235 16 .29
150 LZS 0.01 9.37 25.19 201 .942 -1.8 -17.4 1.48 0.92 3.81 53 19 .14 630 8 1.08 8.36 4 2.33 200 16 .25
151 (11.88) 6.0 11.92 29.00 164 .958 -0.7 -9.1 2.31 2.85 428 150 19 .41 813 8 1.50 884 4 2.54 210 16 .26
95524262
Table 13. Sensitivity Analysis of Little Chestuee (Part 4 of 4)
SENSITIVITY ANALYSIS FOR. LITTLE CHESTUEE WATERSHED WATER YEAR 1981 DRAINAGE ARF4 8.24 Square Miles
DAILY STREAMFLOW STATISTICAL SUMMARY STORM ANALYSIS SUMMARY
STORM OF STORM OF STORM OF STORM OF
FLOW ERROR AVERAGE LOW FLOW 1/14 / 51 2 / 1 /51 3/29 / 51 9' 22 / 51STATISTICS STATISTICS ERRORS STATISTICS #1 #2 #3 #4
ROOT OAILY MEAN MEAN LOWEST PEAK R/O PEAK R/O PEAK R/O PEAK R/OMEAN STO SUM CORR LOW S IGW OF OF OFEV. SQUARE COEFF. FLOW F33.11 (Aug. June Feb. ICFS) (HOUR) I N ICFSI (HOURII IN I (CFS) IOUR) I N I (CFS) IOU : IN
OBSERVEDCONFIGURATION 14.01 27.9 1 5.25 7.24 11.0 191 21 0.49 798 9 1.61 954 2 2.25 230 15 .20BEST SIMULATEDCONFIGURATION 14.78 33.67 194 .963 +0.7 +1.6 4.50 10.97 9.7 222 19 0.61 950 8 1.86 ' 954 4 2.83 231 16 .28
ID PARAM VALUE STARTING MOISTURE STORAGE VALUES AS OF OCTOBER 1 (CONTINUED)
154 18.0 16.83 41.82 345 .946 +0.5 +20.0 10.05 8.22 4.84 288 19 .83 1092 8 2.32 1075 4 3.32 315 16 .39155 24.0 17.72 41.90 347 .948 +0.5 +24.6 10.36 8.29 5.12 231 19 .83 1007 8 2.29 1023 4 3.19 483 16 .59156 BNFX 0.01 14.78 3367 194 .963 +0.7 +1.6 5.63 7.13 4.49 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28157 Fl0747 0.40 14.78 33.67 194 .963 +0.7 +1.6 5.63 7.13 4.49 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
159 .400 14.78 33.67 194 .963 +0.7 +1.6 5.64 7.13 4.49 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
162 IFS 0.1 14.84 33.66 195 .962 +0. +1.6 6.35 7.13 4.49 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28164 (0.0) 0.5 15.08 33.89 213 .954 +1.4 +1.6 9.21 7.13 4.49 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
165 1.0 15.38 34.77 264 .927 +2.0 +1.6 12.79 7.13 4.49 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28167 5.0 17.81 57.81 931 .544 +7.4 +1.6 41.37 7.13 4.49 222 19 .61 950 8 1.86 954 4 2.83 231 16 .28
EVAPORATION DATA
106 EPAET 19.14 18.44 37.79 304 .935 +2.6 +18.6 8.60 16.63 10.4 246 19 .70 998 8 2.03 993 4 3.07 566 16 .70108 131.9) 25.52 16.35 35.40 234 .954 +1.6 +9.1 5.98 13.45 10.0 235 19 .65 974 8 1.95 973 4 2.94 377 16 .46111 38.28 13.48 32.18 169 .966 -0.1 -4.7 3.57 9.24 9.4 210 19 .56 927 8 1.79 935 4 2.71 139 16 .18113 44.66 12.35 30.83 157 .967 -0.7 -10.8 2.89 7.93 9.0 196 19 .51 904 8 1.71 915 4 2.62 75 16 .10114 MNRD 81.0 15.60 34.98 220 .958 +1.1 46.2 5:10 12.18 9.7 233 19 .64 973 8 1.93 974 4 2.92 303 16 .37116 (135.0) 108.0 15.17 34.30 207 .961 +0.9 +3.8 4.77 1154 9.7 228 19 .62 961 8 1.90 964 4 2.87 263 16 .32119 162.0 14.40 33.06 183 .964 +0.5 -0.5 4.25 10.46 9.7 217 19 .59 938 8 1.83 945 4 2.78 200 16 .25121 189.0 14.04 32.46 174 .966 +0.3 -2.5 4.01 10.00 9.7 212 19 .57 927 8 1.80 935 4 2.73 170 16 .21
955-2426.2
Table 14. Summary of Sensitivity Analysis for Little Chestuee(Part 1 of 3) %Nb
PARAMETERS CHANGE IN RESPONSE TO PARAMETER VARIATION
I3 MAJOR WINTER STORM 3/29151 e4 MAJOR SUMMER STORM 9/225 MEAN DAILY FLOW
PAAMTR AE N PEK RIO PE lS RiO TOTAL LOW DRIEST MO./Aug.CHNG U CNG UIS . UI CA UIS CHANEUIS UIS
WATERSHED PARAMETERS
51 AREA -50 -50 +1.00 -50 +1.00 -50 .1.00 -50 +1.00 -50 +1.00 -50 +1.00
52 (8.24) +50 +50 +1.00 50 +1.00 +50 +1.00 +54 +1.07 +50 +1.00 +50 +1.00
53 FIMP -100 -0.42 +0.004 -1.05 +0.010 -6.5 +0.065 -7.14 +0.071 -1.3 +0.013 -3.1 +0.031
57 10.013) +300 +1.57 +0.005 +2.47 +0.008 +18.2 +0.061 +25.0 +0.083 +3.9 +0.013 +8.9 +0.030
FIMP
58 0.10 -3.5 +5.6 +41 +54 +8.7 +20
59 0.20 +7.2 .+13 +89 +111 +19 +43
60 0.50 +18.9 +33 +235 .289 +49 +112
FWTR
63 FWTR 0.01 +0.42 +0.35 +3.9 +7.1 -0.7 -9.1
65 10.0) 0.10 +3.67 +5.65 +42 +50 -3.2 -26
67 0.50 +18.2 +29.6 +210 +250 +23 +51
.... SOIL MOISTURE PARAMETERS 2 0
2 BMIR 50 +15 -0.30 +19 -0.38 +54 -1.08 +54 -1.08 +3.6 -0,07 -24 +0.48
3 (6.3) +50 -14 -0.28 -16 -0.32 -30 -0.60 -25 -0.50 -1.7 -0.03 +13 +0.26
6 LZC -50 +09 -0.18 +14 -0.28 -14 +0.28 -11 +0.22 +13 -0.26 -05 +0.09
7 (8.0) +50 -10 -0.20 -14 -0.28 +19 +0.38 +21 +0.42 -15 -0.30 0.0 0.0
10 ETLF -50 +1.0 -0.02 +0.7 -0.01 +82 -1.64 +82 -1.64 +4.3 -0.09 +15 -0.30
11 (0.25) +50 -0.1 0.0 -0.4 -0.01 -39 -0.78 -36 -0.72 -1.6 -0.03 -5.8 -0.12
14 SIAC -50 -05 +0.10 -06 +0.12 +11 -0.22 +11 -0.22 -0.7 +0.001 +2.0 -0.04
15 (0.30) +50 +04 +0.09 +05 +0.10 -11 -0.22 -07 -0.14 +0.10 +0.002 -2.2 -0.04
18 SUZC -50 +0.0 0.0 -0.4 +0.01 +42 -0.84 +43 -0.86 +3.7 -0.07 420 -0.40
19 (0.30) +50 +0.5 +0.01 +0.4 +0.01 -40 -0.81 -36 -0.71 -3.1 -0.06 -11 -0.22
22 BUZC -50 0.0 0.0 -0.35 +0.01 +5.6 -0.11 +7.1 -0.143 +0.2 -0.004 +0.6 -0.01
23 (0.42) +50 +0.1 0.0 0.00 0.0 -6.5 -0.13 -3.6 -0.70 -0.2 -0.004 -0.9 -0.02
40 BIVF -50 +1.9 -0.04 +1.4 -0.03 0.0 0.0 +3.6 -0.071 -0.07 +0.001 +0.2 -0.004
41 (040) +50 -5.7 -0.11 -4.6 -0.09 -0.87 -0.02 0.0 0.00 -0.30 +0.006 -2.9 -0.058
68 VINTMR -100 +0.1 -0.001 -1.1 +0.01 -0.4 +0.004 +3.6 -0.036 +0.7 -0.007 +5.0 -0.050
69 10.149) -50 0.0 0.00 -0.4 +0.01 -0.4 +0.009 +3.6 -0.072 +0.1 -0.001 +0.9 -0.018
71 +50 +0.1 +0.002 +0.4 +0.01 0.0 0.00 00 0.00 -0.1 -0.002 -0.9 -0.018
72 +100 +0.1 +0.001 +0.7 +0.01 0.0 0.00 0.0 0.00 -0.2 -0.002 -1.6 -0.016
74 +571 +0.8 +0.001 3.2 +0.01 +1.3 +0.002 +3.6 +0.006 -0.9 -0.002 -6.7 -0.012
SUBWF
122 SUBWF 0.05 0.00 0.00 -0.35 -0.070 -0.43 -0.086 0.00 0.00 -2.7 -0.540 -4.9 -0.98
123 (0.0) 0.10 -0.10 -0.01 -0.71 -0.071 -0.43 -0.043 0.00 0.00 -5.5 -0.550 -9.6 -0.96
124 0.20 -0.31 -0.02 -1.05 -0.052 -0.43 -0.022 0.00 0.00 -11 -0.545 -19 -0.95
125 0.30 0.52 -0.02 -1.41 -0.047 -0.43 -0.014 0.00 0.00 -16 -0533 -29 -0.97
GWETF
126 GWETF 005 0.00 I 0.00 -0.35 -0070 -0.87 -0.174 0.00 000 -8.5 -1.70 -49 -9.8
127 0.0 010 -010 -0.01 -0.71 -0.071 -087 -0087 3.57 -0357 -14 -1.40 -70 -70
128 0.20 0.21 -0.01 -0.71 -0.036 -1.30 -065 3.57 0179 21 1.05 -86 -43
129 0.30 -031 -0.01 -1.05 -0.035 -1.30 -0.043 -3.57 -0.119 -26 0.857 -91 -3.0
3-50
Table 14. Summary of Sensitivity Analysis for Little Chestuee(Part 2 of 3)
WY '51
PARAMETERS CHANGE IN RESPONSE ro PARAMETER VARIATION
S .3 MAJOR WINTER STORM 3129151 .4 MAJOR SUMMER STORM 9122/51 MEAN DAILY FLOW
PARAMTER PEAK TOTAL FLO DRIEST MO.I A
..-. U. I .. N.A U.. S. u' . ... CINHIIE jCI NO o .. i U . ANA uS
.... OVERLAND FLOW PARAMETERS
46 IFRC -50 0.52 -0.010 +0.35i -0007 0.00 0.00 0.G 0.00 0.0 0.00 -0.2 +0.004
47 10.10) 0.31 -0.006 -0.71 -0014 0.00 0.00 00 0.00 0.0 O -0.2 -0.004
77 OFSS -50 +0.42 O0.008 -0.35 +0.007 -0.43 +0.009 0.0 000 -0.2 +0.004 +1.3 -0.026
78 (0.237 +50 -0.10 -0.0020.00 000 0.00 0.00 13.6 10.071 +0.07 +0.001 -0.9 -0.018
79 +100 0.00 0.00 0.00 +3.6 +0.036 .0.2 +0.002 -1.3 -0.013
81 +300 -0.31 -0.001 +0.35 +0.001 0.00 000 +3.6 +0.012 '0.3 +0.001 -2.4 -000
82 OFSL -100 -0.31 +0.003 +1.05 -0.010 0.0 0.0 +3.6 -0.036 +0.7 -0.007 -5.3 .0.053
83 1630.0) -50 -0.31 -0.006 +0.35 -0.007 0.0 0.0 +3.6 -0.071 +0.3 -0.007 -2.4 +0.048
84 +50 +0.52 +0.010 -0.35 0.007 -0.9 -0.2 0.0 0.0 -0.3 -0.006 +1.6 40032
85 +100 +084 +0.008 -0.71 -0.007 -6.5 -0.06 -3.6 -0.04 -0.5 -0.005 +29 .0.029
87 1400 +1.26 +0.003 -1.77 -0.004 -29 -0.07 -25 -0.06 -1.2 -0.003 +8.0 +0020
88 OFMN -51 -0.31 +0.006 +0.35 -0.007 0.0 0.0 3.6 -0.07 +0.34 -0.007 -2.4 +0.047
89 (0061) +49 +0.52 +0011 -0.35 -0.007 -0.9 -0.02 0.0 0.0 -0.2 -0.004 +1 6 +0.033
92 +400 +1.36 +0.003 -1.77 -0.004 -29 -0.07 25 -0.06 -1.2 -0.003 +80 +0.020
93 OFMNIS -75 0.00 0.00 0.00 0.00 0.0 0.0 0.0 0 0.0 0 0.00 -0.2 +0.003
97 (0.0021 +400 000 0.00 0.00 0 00 0.0 0.0 0 0 0 0.0 0.00 -0.2 0000
" CHANNEL ROUTING AND GROUND WATER PARAMETERS
28 CSRX -50 +0.8 -0.02 +2.47 -0.05 +35.0 -0.70 +3.6 -0.071 +0.2 -0.004 0.0 0.00
29 (0.95) +50 -22 -0.44 -12.0 -0.24 -44.6 -0.89 -10.7 -0.214 -0.1 -0.002 -0.4 -0.01
32 FSRX -50 0.0 0.00 0.00 0.00 0.00 0.0 0.00 0.00 0.0 0.00 00 +0.004
33 (0.95) +50 -18 -0.36 -1.77 -0.04 -0.43 -0.01 +3.6 +0.071 +0.3 +0.006 -0.2 -0.004
36 BRFC -50 +0.4 -001 +0.35 -0.01 -0.43 +0.01 0.00 0.00 +1.3 -0.026 -30 +0.6037 (0.98) 150 -0.5 -0.01 -1.41 -0.03 +0.43 -0.01 17.14 +0.14 -6.2 -0.124 +47 +0.94
99 CHCAP -75 0.0 0.00 0.0 00 00 000 0.0 00 00 0.00 -0.2 _0.003
103 (200.0) '50 0.0 0.00 0.0 0.00 0.0 0.00 0.0 0.0 0.0 0.00 -0.2 -0.004
104 +100 0.0 0.00 0.0 0.00 0.0 000 0.0 0 0 0.0 0.00 -0.2 -0.002
130 EXOPV -97 00 0.00 0.0 0.00 0.0 0.00 00 0.0 0.0 0.00 -0.2 0.002
131 (0.301 -50 0.0 0.00 0.0 0.00 0.0 0.00 0.0 0.0 0.0 0.00 -0.2 +0.004
132 +50 0.0 0.00 00 0.00 000 0.00 0.0 0.0 0.0 0.00 -0.2 -0.004
134 +233 0.00 0.0 0.00 0.0 0.00 0.0 0.0 0.0 0.00 -0.2 -0.001
135 FNLR 88 +0.31 -0.004 +0.35 -0.004 -0.43 0.005 00 0.0 +034 -0.004 -13 0.147
136 (0.85) -65 +021 -0.003 0.00 0.000 -0.43 +0006 00 0.0 0.14 -0002 7.1 +0.110
13-7 41 40.10 -0.002 000 0.000 -0.43 *0,010 0.0 0.0 +0.07 -0.002 3.8 0.092
138 18 10.10 -0.006 0.00 0.000 0.00 0.00 0,0 0.0 0.00 0.00 1.6 +0.091------ ... . .. .
139 .9 0.00 0.00 -0.35 -0059 0.00 0.00 0.0 0.0 000 0.00 0.2 +0.034
3-51
Table 14. Summary of Sensitivity Analysis for Little Chestuee
(Part 3 of 3)WY '51
PARAMETERS CHANGE IN RESPONSE TO PARAMETER VARIATION
3 MAJOR WINTER STORM 3/29151 4 MAJOR SUMMER STORM 9/22/51 MEAN DAILY FLOWID PARAMETER APERENT PEAK R PEAK R/O TOTALFLOW DRIEST MO./Aug.
VARIATION HANGE %CHANGE U %XCHANGE U/S %CHANGE U/S %CHANGE U/S %CHANGE U/S
- STARTING MOISTURE STORAGE VALUES AS OF OCTOBER 1 -
141 GWS -97 0.00 0..0 0 0.00 0.00 .00 0.00 -1.4 +0.014 +3.3 -0.034
142 (0.355) -49 0.00 .000 0.00 0.00 0.00 0.00 -0.7 +0.014 0.0 0.00
143 +52 0.00 0.00 0.00 0.00 000 0.00 0.00 0.00 +0.7 +0.013 +3.3 +0.063
144 +100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 +1.4 +0.014 +3.3 +0.033
UZS
145 UZS 0.1 +0.21 0.0 0.00 +3.6 +0.2 +3.6
146 (0.0) 0.5 +0.84 +1.05 +0.87 +3.6 + +1.2 +4.0
147 1.0 +1.15 +1.41 +1.73 +3.6 +3.0 +0.9
150 LZS -100 -12.4 +0.124 -17.7 +0.177 -13.4 +0.134 -10.7 +0.107 -37 +0.37 -15.0 +0.150
151 (11.88) -50 -7.3 +0.148 -10.2 +0.205 -9.1 +0.184 -7.14 0.144 -19 +0.38 -4.7 +0,095
154 +52 +12.6 +0.245 +17.3 +0.336 +36.4 +0.707 +39.3 +0.763 +14 +0.27 +7.8 +0.151
155 +102 +7.2 +12.7 +0.125 +109 +1.07 +111 +1.09 +20 +0.20 +14 +0.137
156 8FNX -99 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.0 0.00 0.0 0.00
157 (0.747) -47 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.0 0.00 0.0 0.00
159 +34 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.0 0.00 0.0 0.00
IFS
162 IFS 0.1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 +0.4 +4.0 0.0 0.00
164 (0.0) 0.5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 +2.0 +4.0 0.0 0.00
165 1.0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 +4.1 +4.1 0.0 0.00
157 5.0 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 +16.4 +3.3 0.0 0.00
.... EVAPORATION DATA
106 EPAET -40 +4.1 -0.10 +8.5 -0.210 +145 -3.62 +150 -3.75 +25 -0.62 +91 -2.28
108 (31.9) -20 2.0 -0.10 +3.9 -0.190 +63 -3.15 +64 -3.20 +11 -0.55 +33 -1.65
111 +20 -2.0 -0.10 -4.2 -0.210 -40 -2.00 -35 -1.75 -8.8 -0.44 -21 -1.05
113 +40 -4.1 -0.10 -7.4 -0.186 -68 -1.70 -64 -1.60 -16 -0.40 -36 -0.90
114 MNRD -40 +2.1 -0.05 +3.2 -0.080 +31 -0.775 +32 -0.80 +5.5 -0.14 +13 -0,325
116 (135.0) -20 +1.1 -0.05 +1.4 -0.075 +14 -0.695 +14 -0.72 +2.6 -0.13 +6.0 -0.300
119 +20 -0.9 -0.05 -1.8 -0.088 -13 -0.670 -11 -0.54 -2.6 -0.13 -5.6 -0.280
121 +40 -2.0 -0.05 -3.5 -0.088 -26 -0.660 -25 -0.62 -5.0 -0.12 -11 -0.275
3-52
3.6 CORRELATION (TASK 6)
The data base for the Correlation task consists of (1) the ten sets
of "best" model parameters from the Calibration task and (2) ten sets of
Watershed Parameters (i. e., observable or inferable characteristics)from the Physiographic Data Base Task, as indicated in Figure 28. Theobjective of the task was to derive a statistically based set of formulasfor quantifying those parameters which are normally estimated by thecalibration process previously described.
A linear correlation coefficient was calculated for each of 88pairs of parameters. In each pair, one parameter was chosen fromthe set of calibrated model parameters and the other from the set of"observable" parameters. These correlation coefficients, for a linearrelationship (y = a + bx), were found to be generally rather low, theexceptions (up to 0.8) being associated with parameters whose valueslie in narrow ranges. The results of this step provided a basis forselection of combinations of observable parameters for the multi-variate linear regression analysis. It was necessary to accept linearcorrelation coefficients as low as 0.39 in order to generate a useableexpression for each calibrated parameter.
The result of the multivariate regression analysis 1 6 is the setof equations listed in Table 15. They were used to estimate modelparameters for the Validation task. The parameters FIMP, OFSS,OFMN, OFSL, and OFMNIS can be determined from remotely sensedimage data, provided resolution (not yet determined precisely) is ade-quate and relative elevations can be discerned. The soil parametersAWC (Available Water Capacity) and PERM-A (Permeability of the AHorizon) were obtained from County Soil Surveys of the U.S. Soil Con-servation Service; they are not remotely sensible at present. In afuture application in an ungaged area, some ground truth data relatedto soil characteristics will be required unless t ose characteristicscan be inferred from surrogate information ,-
The data base sample points are widely scattered in all cases.This is evident from the statistical indices calculated and shown inTable 16. It would obviously be worthwhile to refine the CorrelationAnalysis. Additional observables should be introduced, carefully chosenwith regard both to hydrologic phenomena and remote sensing systemcapabilities. Additionally, interaction between independent variablesshould be resolved through translation into an intermediate set of com-posite variables. Nevertheless, this simplified analysis produced re-sults which led to acceptable validation runs and a strong indication ofthe feasibility of the technique.
3-53
OBSERVABLE ANDINFERABLE WATERSHEDCHARACTERISTICSFROM TASK 1
' TO TASKS 7, 8, 9
ANALYSISMODELPARAMETERS PARAMETER-PHYSIOGRAPHYFROM TASK 4 RELATIONSHIPS
Figure 28. Task 6: Correlation
Table 15. Correlation Results (Multiple Regression Analysis)
BMIR = 5.27 + 0.75 PERM A
LZC = 1.80 + 4.73 OFSS + 2.63 AWC + 0.26 PERM A
ETLF = 0.05 + 1.49 OFMN
SUZC = 0.46 + 2.51 FIMP - 0.04 AWC
SIAC = 0.80 + 3.84 FIMP - 0.02 AWC - 0.37 OFSS - 0.0003 OFSL
c BUZC = 0.18 + 9.69 OFMN - 0.0003 OFSLIc
BIVF = -0.02 - 9.25 FIMP + 0.0004 OFSL
CSRX = 0.95 + 1.92 OFMNIS - 0.003 AWC
FSRX = 1.01 - 0.006 PERM A - 0.00003 OFSL
BFRC = 0.97 - 3.29 FIMP
BFNLR = 0.86 - 4.38 FIMP
Table 16. Statistical Characteristics, Regression Equations
PARAMETER S. M. S.E.E. S.M. C. C.
BMIR 6.94 2.89 0.15
LZC 4.96 1.26 0.58
ETLF 0.135 0.060 0.17
SUZC 0.290 0.153 0.42
SIAC 0.300 0.227 0.47
BUZC 0.349 0.109 0.82
BIVF 0.296 0. 332 0.36
CSRX 0.957 0.016 0.50
FSRX 0.959 0.022 0.43
BFRC 0.935 0.088 0.24
BFNLR 0.815 0.097 0.31
S. M. = Sample Mean
S.E.E. = Standard Error of Estimate
S. M. C. C. = Square of Multiple Correlation Coefficient
3.7 VALIDATION (TASK 7)
Five basins of the Tennessee Valley were identified for theValidation task. They were exempt from calibration and treated as"ungaged" watersheds. The watershed description and historicaldata from the integrated data base, and the model parameters esti-mated by the linear regression equations generated in the Validationtask, are inputs, as indicated in Figure 29. For each basin, a multiyearsimulation was run, the synthesized streamflow results compared withobserved streamflow, and analyses of the results printed out in tabularand plot formats.
Results of the validation runs are summarized in Table 17. Meanmonthly streamflow correlation coefficients indicate generally good agree-ment between simulated and observed values. For each basin, the largestwinter storm event was chosen for comparison of simulated streamflowpeaks with observed peaks.
Examples of daily and hourly plots are shown in Figures 30 and31 respectively. In the latter, the observed storm event is representedonly by the flood peak. Peak magnitude and timing are the basis forjudging the effectiveness of the model with respect to storm events.Plotting an observed runoff hydrograph at one-hour intervals wouldrequire a prohibitively expensive manual translation of strip-chartrecords into hourly runoff for every selected storm event.
The result of the Validation task is a strong indication of thefeasibility of making streamflow forecasts for an ungaged basin, atleast with respect to mean daily streamflow. With respect to stormrunoff peaks, the result is inconclusive. Further refinement of thetechnique will lead to the ultimate method depicted in Figure 32.
3.8 RESOLUTION/ACCURACY ANALYSIS (TASK 8)
This task was not performed in the study because of budgetarylimitations. It is being addressed in a separate NASA study, the resultsof which will be available in the second quarter of 1974.
3-57
WATERSHED CHARACTERISTICSFROM TASK 1
SIMULATED STREAMFLOW
MODEL
PARAMETER MODEL SIMULATION OBSERVED STREAMFLOW EVALUATIONPARAMETER VALUE
co QUANTIFICATION STATISTICAL INDICESco KWM
PHYSIOGRAPHY-PARAMETER DATARELATIONSHIPS: FROM TASK 6 FROM TASK 3 Prediction Accuracy
_ TO TASK 9
EVALUATION OFPREDICATIONACCURACY
Figure 29. Task 7: Validation
Table 17. Validation Results
STREAMFLOW LARGEST WINTER STORMWATERSHED YEARS
OF CORRELATION DAILY, CFS , PEAK
AREA RECORDS COEFFICIENT MEAN MAX DATE FLOW, CFS TIME, HRNO. SHORT NAME
MI2 KM 2 YRS FROM THRU M'LY D'LY OBS SIM OBS SIM OBS SIM O3S SIM
4 NOLAND CREEK 13.8 35.7 9 56 64 .95 .83 47.2 37.7 923 982 12-12-61 1,371 1,375 1 1
ci1 6 BIG BIGBY CREEK 17.5 45.3 5 64 68 .92 .86 27.0 24.4 1020 1707 3-6-67 1,510 1,069 40 35
8 LITTLE RIVER 26.8 69.5 5 65 69 .96 .87 108 92 2840 3675 2-13-66 4,049 7,727 7 6
10 PINEY CREEK 55.8 145 8 61 68 .96 .86 94 98 4220 6242 3-12-63 12,900 13,597 25 28
16 EAST FORK POPLAR 19.5 50.5 8 62 69 .93 .87 47 31 1450 1456 3-12-63 1,802 2,013 20 5
NOLAND CREEK, NC WATER YEAR 19631000 28.3
900 25.5
800 22.8
zoo 700 19.8 00 wC.)
I- SIMULATED STREAMFLOW
S600 - - - OBSERVED STREAMFLOW "17.0w" 600 17.0OBSERVED TOTAL = 4,505
I SIMULATED TOTAL = 3,817LU
U- I I 1
500 14.2
0I 0
200 , 5.7
lo .I100 8 cn . . . ..
0.0 0.04 8 12 16 20 24 28
DAYS IN MARCH
Figure 30. Example of Mean Daily Streamflow Plots
3-60
NOLAND CREEK, NC 03/05/63 37.51300 37.5
OBSERVED FLOOD PEAK
1200 I 34.6
\ - SIMULATED STREAMFLOW\ -- -- -- OBSERVED PRECIPITATION
1100 (UNSCALED) 31.7
OBSERVED RUNOFF = 1.51 IN. (SIMULATED RUNOFF = 1.83 IN. (
1000 - 28.31000II 0
z 900 26.0 UO 0
U 800 23.1 "
w 20.2uj
w 700 20.2U. I
S600 17.3
oI 0S500 14.4 -jIII
LU 1 5
- 400 11.5
300 8.6
200 5.7
100 2.8
0.0 0.00.0 10 20 30 40
TIME, HOURS
Figure 31. Example of Hourly Streamflow Plots
3-61
PRECIPITATION
OBSERVABLE
CHARACTERISTICS
CATALOG
CATALOGMODEL SIMULATION STREAMFLOWUNGAGED INFERABLE MODEL -
U GB OF PARAMETERSPREDICTIONBASIN CHARACTERISTICS MODEL
DATA RELATIONSHIPS
A PRIORI
INFORMATION
Figure 32. Ultimate Method
SECTION 4
RECOMMENDED ADDITIONAL RESEARCH
As in all such research, this feasibility study revealed severaltopics deserving further study. Also, the investigators believe thatsome tasks should be done more thoroughly and/or more efficientlywith the benefit of experience gained in subsequent tasks. The follow-ing paragraphs summarize topics the authors recommend for additionalresearch.
4.1 EXPANDED CORRELATION DATA BASE
The five validation runs used data derived from the results of acorrelation analysis which was in turn based on calibration of only tenbasins. The size of the sample population limits confidence in the re-sults; the good correlations scored in the validation runs may well havebeen "good luck". It is suggested that the following tasks be performed.
* Collect historical and physiographic data for twenty additionalwatersheds of the Tennessee Valley, including five from theMississippi Alluvial Plain Region.
* Preprocess and format the new data and integrate it intothe Master Data Bank.
* Rework the Correlation Analysis using the expanded database (15 basins).
* Quantify parameters for ten basins, the five used for vali-dation in the basic study and five reserved for the purposefrom the new set.
* Perform ten validation runs and analyze the results.
* Revise this report or issue a supplementary report.
4.2 REFINED CORRELATION STUDY
Even with the limited population sample (ten sets of parameters)used in the basic feasibility study, the correlation analysis should beredone with a more rigorous methodology to achieve more refined re-sults. It is desirable, for instance, to remove the sub-surface aspects,such as permeability and water holding capacity, because they are notdirectly measurable or inferrable from remote sensing. Additionalphysical characteristics which lend themselves to determination by
4-1
remote sensing, even from satellite altitudes, should be introducedinto the study... Hydrologic expertise should be brought to bear in theselection of combinations of observable characteristics. Intermediateor composite variables should also be derived in order to eliminate theeffects of interaction among the selected independent parameters. Thisadditional research task should produce a report which is supplementaryto this one.
4.3 VALIDATE TECHNIQUE IN OTHER REGIONS
A similar study could be performed in another region which isphysiographically and climatically very different from the TennesseeValley, as a further validation of the feasibility of the technique. Thisadditional work should concentrate in an area in which snowfall is asignificant form of precipitation.
4.4 DOCUMENT THE SYSTEM FOR SIMULATION AND ANALYSISOF WATERSHEDS
The system for simulation and analysis of watersheds, developedby IBM around the KWM and OPSET programs, is an effective tool forresearch and hydrologic modeling, river forecasting, and water re-sources management. However, it is presently useful only to thosewho are involved in the development of the system, who are familiarwith its operation. Inadequate documentation exists to transfer thetool to another investigating team without an extensive period of learn-ing and familiarization. It is suggested that the system for simulationand analysis of watersheds be documented by the preparation of operatinginstructions and users' manuals containing flowcharts, program descrip-tion, and step by step instructions of how another user could take advantageof this system of programs and methods.
4.5 PILOT DEMONSTRATION, OPERATIONAL RIVER FORECASTAPPLICATION
Accepting the results of the study as a strong indication of thefeasibility of the technique the study originally set out to investigate,it appears appropriate for NASA to leave further refinement of thetechnique to potential users and other researchers. A more appro-priate step for NASA to take may well be pursuit of a study jointlywith a potential user agency, aimed at implementation of a pilotdemonstration of operational river forecasting utilizing the models,methods, and other results of the feasibility study. The first productof such an undertaking would be a specification for the demonstrationsystem, to be followed by implementation of the pilot system. For aperiod of time the pilot system could be operated in parallel with theuser agency's normal river forecast methods. Then the two systems
4-2
could be compared with respect to accuracy and efficiency. Should thedemonstration prove successful, complete documentation of the system,as a NASA product available to potential user agencies, would be thefinal task.
4-3
APPENDIX A
SUMMARY OF STUDY HISTORY
The feasibility study reported herein began in mid-1971. At the
beginning of 1972, it was reduced to a low-level sustaining effort andthen resumed at an appropriate manning level in July 1972. It isappropriate to summarize those three segments of the study in thisAppendix.
A. 1 LAST HALF, CALENDAR YEAR 1971
During the months July through December 1971, with the support of
the Research Institute of the University of Alabama in Huntsville, IBM
* Selected a family of 55 Tennessee Valley Watersheds, withthe advice and assistance of the Tennessee Valley Authority(TVA)
* Collected 30 percent of historical data needed for the study
* Acquired 80 percent of the applicable topographical maps and
photographic coverage
* Programmed and operated the statistical rainfall-runoffcorrelation technique sometimes referred to as the "APIMethod"
* Programmed and operated three basic synthetic hydrographregeneration methods, those of Clark, Nash, and Decoursey,with two additional variations on the Nash method
* After a review of available simulation models, selected theKentucky Watershed Model and its companion calibrationprogram for continuation of the study.
* Established a mutually beneficial technical liaison with theTennessee Valley Authority and orally reported the studyconcept and progress to other organizations: World Bank;U. S. Army Corps of Engineers; U. S. Departments ofInterior (Geological Survey), Agriculture (AgriculturalResearch Service), and Commerce (National Oceanic andAtmospheric Administration); and National Academy ofSciences.
A-1
A. 2 FIRST HALF, CALENDAR YEAR 1972
The initial funding authorization expired at the end of 1971, and thestudy was reduced to a sustaining effort funded at a one-man level fromJanuary through June 1972. Highlights of this interim effort were asfollows:
* The Kentucky Watershed Model and its optimization programwere acquired and operated in the IBM-Huntsville facility.Additionally, the Tennessee Valley Authority Daily FlowModel, an experimental simulation program, was acquired,set up, and run.
* Modifications and improvements in the KWM and OPSETwere begun during this interim period in order to adapt itto the purposes of the feasibility study. These additionsand improvements included preprocessing routines forinput data and implementation of several plot and tabularoutput options.
* The historical data base was expanded by the acquisitionof digital daily streamflow tapes from USGS and digitalprecipitation data tapes from the National Weather Service.In order to provide hourly precipitation data for the masterdata bank, a computer program to merge and synchronizerainfall data from several hourly and daily precipitationgages was completed and made operable.
A. 3 SECOND HALF, CALENDAR YEAR 1972
In July 1972 the study was resumed at a five-man funding level,which was reduced to three in mid-September. Highlights of the lastsix months of 1972 are as follows:
* The design of an integrated data bank consisting of historicaldata and model parameters derived from the physiographicdata base was completed. A master data tape was establishedand data pertaining to three TVA watersheds were entered onthe tape. The master data tape has the capacity for storageof up to 50 watersheds, with data related to 10 years of historicalexperience for each one.
* The calibration program and simulation model were testedusing actual data from TVA watersheds. After calibrationthe simulation model is capable of synthesizing streamflowswhich are accurate within five percent of observed flows.Accurate simulation of flood peaks with respect to magnitude
A-2
and time of peak require additional manual adjustments aftermodel parameters have been estimated by OPSET.
* A sensitivity analysis task was introduced into the study inorder to determine what effects on accuracy of simulationvariations in the different model parameters have. Thisanalysis was performed on two small watersheds of thetest area. It helped expedite calibration of the remainingwatersheds.
* The data base and models were implemented in the IBM facilitywith a Terminal Assistance Package which allows an operatorto set up programs, modify the data base, and initiate programoperations from a terminal without having to handle severalthousands of data cards.
A. 4 CALENDAR YEAR 1973
The study was continued at a three-man funding level from January 1through September 21, 1973, and was completed, with the results assummarized in Section 1. 5 and detailed in Section 3. Most of the effortduring this time was spent on formatting and integrating the data base for15 watersheds. Calibration of ten watersheds required the next largestproportion of time, after which the correlation analysis and validation runswere quickly completed.
A-3
APPENDIX B
REFERENCES
1. Castruccio, P. A., "Use of Remote Sensing in Hydrology, "XXII International Astronautics Congress, Brussels, 1971.
2. Linsley, Ray K., Kohler, Max A., and Paulhus, Joseph L. H.Hydrology for Engineers. New York: McGraw-Hill Book Co.,1958.
3. Thomas, D. M., and Benson, M. A. "Generalization of Stream-flow Characteristi cs from Drainage Basin Characteristics, " OpenFile Report, U. S. Department of Interior, Geological Survey, 1969.
4. Crawford, Norman H., and Linsley, R. K. Digital Simulation inHydrology: Stanford Watershed Model IV. Stanford, California:Stanford University, Department of Civil Engineering, TechnicalReport No. 39, July, 1966.
5. Hydrocomp International. "HSP Operations Manual, " Palo Alto,California, 1969.
6. Anderson, E. A. and Crawford, N. H. The Synthesis of ContinuousSnowmelt Hydrographs on a Digital Computer. Stanford, California:Stanford University, Department of Civil Engineering, TechnicalReport No. 36, 1964.
7. Hydrocomp International. "Simulation of Continuous Discharge andStage Hydrographs in the North Branch of the Chicago River, "Palo Alto, California, 1970.
8. James, L. Douglas. Economic Analysis of Alternative Flood ControlMeasures. Lexington: University of Kentucky, Water ResourcesInstitute, Research Report No. 16, 1968.
9. Ligon, James T., Law, Albert G., and Higgins, Donald H.Evaluation and Application of a Digital Hydrologic Simulation Model.Clemson, South Carolina: Clemson University, Water ResourcesResearch Institute, Report No. 12, November, 1969.
10. Lumb, Alan Mark. Hydrologic Effects of Rainfall Augmentation.Stanford, California: Stanford University, Department of CivilEngineering, Technical Report No. 116, November, 1969.
B-1
REFERENCES (Continued)
11. James, L. Douglas. An Evaluation of Relationships between Stream-flow Patterns and Watershed Characteristics through the Use of OPSET:A Self-Calibrating Version of the Stanford Watershed Model.Lexington: University of Kentucky, Water Resources Institute,Research Report No. 36, 1970.
12. Liou, Earnest, Y. OPSET: Program for Computerized Selection ofWatershed Parameter Values for the Stanford Watershed Model.Lexington: University of Kentucky, Water Resources Institute,Research Report No. 34, 1970.
13. Ross, Glendon A. The Stanford Watershed Model: The Correlationof Parameter Values Selected by a Computerized Procedure withMeasurable Physical Characteristics of the Watershed. Lexington:University of Kentucky Water Resources Institute, ResearchReport No. 35, 1970.
14. Kirpich, Z. P., "Time of Concentration of Small AgriculturalWatersheds," Civil Engineering, Vol. 10, June 1940, p. 362.
15. Chow, Ven Te (ed.). Handbook of Applied Hydrology. New York:McGraw-Hill Book Company, 1964.
16. Snedecor, G. W., and W. G. Cochran, Statistical Methods, SixthEdition, Ames, Iowa: The Iowa State University Press, 1967.
17. Root, R. R., and Miller, L. D. Identification of Urban WatershedUnits Using Remote Multispectral Sensing. Fort Collin, Colorado:Colorado State University Environmental Resources Center,Completion Report Series No. 29, June 1971.
18. U. S. Department of Agriculture, "Aerial-Photo Interpretation inClassifying and Mapping Soils, " Soil Conservation Service Agricul-ture Handbook 294, October 1966.
B-2
APPENDIX C
PROGRESS REPORT BASIC DISTRIBUTION LIST
IBM-HUNTSVILLE
R. Ambaruch, 53-K31C. L. Martin, 13-K36N. R. Ruest, 53-D54J. A. White, 13-K24 (7)J. W. Simmons, 53-K31Library, 53-U34 (4)
IBM- GAITHERSBURG
R. Bernstein, 2T01FL2S. Shapiro, 1P50FL3P. G. Thomas, 1C41GD8
IBM-HOUSTON
Dr. D. S. Ingram, Mail Code 58
IBM- FSD MARKETINGSuite 9201120 Connecticut Ave., NW.Washington, D. C. 20036
A. Adelman (7)
IBM-THOMAS J. WATSON RESEARCH CENTERP.O. Box 218, Rt. 134Yorktown Heights, N. Y. 10598
Dr. C. K. ChowDr. R. A. FreezeDr. J. S. SmartDr. J. R. Wallis
C-1
COLORADO STATE UNIVERSITYFort Collins, Colorado 80521
Dr. G. R. Johnson, University Computer CenterDr. W. D. Striffler, Department of Earth ResourcesDr. L. D. Miller, Department of Water Sciences
GEORGIA INSTITUTE OF TECHNOLOGYEnvironmental Research CenterAtlanta, Georgia 30332
Dr. L. D. James, Associate Professor (2)
GEORGIA INSTITUTE OF TECHNOLOGYSchool of Civil EngineeringAtlanta, Georgia 30332
Dr. J. R. Wallace (2)
UNIVERSITY OF ALABAMA IN HUNTSVILLEP. O. Box 1247Huntsville, Alabama 35807
Dr. J. J. BrainerdDr. Ken Johnson, Center for Environmental StudiesDr. C. C. Shih
NATIONAL ACADEMY OF SCIENCEU.S. Committee for the International Hydrologic Decade2101 Constitution Avenue, NW.Washington, D.C. 20418
Dr. Leo Heindl, ChairmanDr. Henry Santeford, Staff Associate
TENNESSEE VALLEY AUTHORITY350 Evans BuildingKnoxville, Tennessee 37902
Alfred J. Cooper Colby V. ArdisRoger P. Betson Donald NewtonClaude H. Smith Russell TuckerDon Eklund Norris Cline
C- 2
NATIONAL AERONAUTICS AND SPACE ADMINISTRATIONGoddard Space Flight CenterGreenbelt, Maryland 20771
Dr. V. V. Salomonson, 650 (2)
NATIONAL AERONAUTICS AND SPACE ADMINISTRATIONGeorge C. Marshall Space Flight CenterHuntsville, Alabama 35812
John Bensko S&E-SSL-S Building 4481W. D. Clarke. S&E-S/P-EK Building 4610B. C. Hodges S&E-COMP-SD Building 4654G. F. McDonough (5) S&E-EA-DIR Building 4202J. C. Sloan S&E-AERO-YT Building 4610W. W. Vaughan S&E-AERO-Y Building 4610Dr. F. R. Krause S&E-AERO-T Building 4610
NATIONAL AERONAUTICS AND SPACE ADMINISTRATIONMississippi Test FacilityBay St. Louis, Mississippi 39520
Dr. J. C. McCall
NATIONAL AERONAUTICS AND SPACE ADMINISTRATIONJohnson Space CenterHouston, Texas 77058
Raymond R. Clemence, HD
NATIONAL AERONAUTICS AND SPACE ADMINISTRATIONLangley Research CenterHampton, Virginia 23365
Robert Parker 477 Building 1202
U.S. DEPARTMENT OF AGRICULTUREHydrography LaboratorySoils BuildingBeltsville, Maryland 20705
Mr. H. N. Holtan, DirectorDr. C. England, Soil Physicist
C-3
U.S. DEPARTMENT OF AGRICULTURE
Soil and Water Conservation Research Division
Southern Great Plains Watershed Research Center
P. O. Box 400Chickasha, Oklahoma 73018
Dr. Donn G. DeCoursey, Director
Mr. Bruce Blanchard
U.S. DEPARTMENT OF AGRICULTURE
U.S. Soil and Water Conservation Research Station
Agricultural Research Service
Athens, Georgia 30601
Dr. Donald Overton
U.S. DEPARTMENT OF COMMERCE
National Oceanic and Atmospheric Administration
National Weather Service, Gramax BuildingOffice of HydrologySilver Spring, Maryland 20910
Dr. Max A. KohlerDr. Eugene PeckMr. John C. MonroMr. Eric A. Anderson
U.S. DEPARTMENT OF COMMERCELower Mississippi River Forecast Center
Slidell, Louisiana 70458
Mr. C. E. Vicroy, Jr.
Mr. C. E. Schauss
U.S. DEPARTMENT OF INTERIORU.S. Geological SurveyP.O. Box VUniversity, Alabama 35486
Mr. Jim DanielMr. C.F. Hains
C-4
U.S. DEPARTMENT OF INTERIOR
U.S. Geological SurveyFederal Building, Room 203
Cullman, Alabama 35055
Mr. Paul Cole
U.S. DEPARTMENT OF INTERIOR
U.S. Geological SurveyP.O. Box 2857Raleigh, North Carolina 27602
Mr. N. O. Thomas
U.S. DEPARTMENT OF INTERIOR
U.S. Geological SurveyResearch Institute BuildingP.O. Box 1247Huntsville, Alabama 35807
Mr. R. C. Christensen
UNIVERSITY OF ALABAMAResearch Institute BuildingP.O. Box 1247Huntsville, Alabama 35807
Mr. F. L. Doyle, ChiefNorth Alabama RegionGeological Survey of Alabama
U. S. ARMY TOPOGRAPHIC LABORATORIESFort Belvoir, Virginia 22060
Mr. Donald Orr
U. S. ARMY ENGINEER DISTRICT601 East 12th StreetKansas City, Missouri 64106
Mr. R. A. Johnson, ED-FG
C-5